Press Release – Instaknow Granted Fifth Patent!

Instaknow’s fifth granted patent (US 9,443,005) demonstrates a new dimension of Instaknow’s Artificial Intelligence in solving complex real-world problems. The latest patent was granted for Instaknow’s new ability of reading and understanding the meaning of natural text, so that simple and complex questions can be answered precisely and automatically.

Emulating Linguistics Intelligence

Branded “InstaNLP®”, the NLP capability can:

• Disambiguate words/phrases in complex sentences
• Take into account conditional statements in the text (If, otherwise, but….)
• Be aware of inferences (e.g. faster speed means less time needed to travel… and all the other inferences we call “Common Sense”)
• Understand the total/combined meaning of multiple sentences/paragraphs/texts (i.e. “Discourse Intelligence”)
• Dynamically look up related text from reliable information sources to provide the best possible answer to a questio

Where can InstaNLP® be used?

Exploding volumes of Data, especially unstructured/plain-text data, is the biggest problem for today’s conventional computing, which requires data to be in a structured format. Using plain text to do computing is a major breakthrough to:

• Fine tune Web and document search results by filtering out irrelevant search results and returning precise answers rather than Web page or document links. Users will be able to ask very precise questions to search engines and find search results matching the precise question, regardless of word, phrase and even language differences. Today no search engine is capable of finding the precise meaning of free form text, to accurately answer precise but complex questions like:
  • “Is the current temperature in Moscow greater than the current temperature in New York?”
  • “How many US presidents died before reaching the age of 80?”
  • “How do I repair the door lock on a 2002 Chevrolet Silverado after it has been in a side-ways crash?”
• Gather finely refined intelligence from text sources. Examples of such sources are Internet/Intranet/Extranet news, articles, blogs, journals, dictionaries, encyclopedias, e-mails, Word, PDF, Excel, documents. E.g. questions like “Who from the U.S. banned group ABC was in London when person-of-interest XYZ was there in 2009?” can be answered automatically. This has dramatic implications for fighting terrorism, human rights violations, corruption, money laundering etc.
• Convert free-form, unstructured information into precise structured data so it can be processed by conventional data processing systems. Commonly found examples are:
  • Automatically read incoming e-mails and instant messages to answer precise customer questions or respond to requests for information.
  • Pulling specific repair instructions and data values out of sentences written in device or procedure manuals and providing those specific instructions to a human user (repair technician) or to another computer
  • Pulling specific treatment instructions and test-result values out of medical transcripts and providing them to nurses, doctors or other computer systems.
  • Match resumes of candidates to specific job position requirements
  • Match patient candidates to clinical trials by comparing structured data from a EMR (Electronic Medical Record) system with clinical trial participation criteria expressed in a natural language (e.g. “A female subject is eligible to participate if she is of non-childbearing potential, and if she is of childbearing potential she must use protocol defined contraception methods.”)
  • Any other usage today requiring manual review, interpretation and input of the relevant part of free form text into a computer system, program or interface
• Make distance learning easier by allowing students to find precise answers, help and guideline with natural language interaction with computer based training, help material and courses.
• Make computer games more interesting by allowing the user to ask questions and provide responses in free form spoken or written text rather than choose from pre-selected options on a screen. E.g. a war-game user can say “If my opponent has more than two ships left, then activate my missile batteries”.
• Control complex machinery, processing systems, computer systems, medical devices, virtual assistants and robots with voice commands spoken or typed in plain English or any other natural language.
• Generate indexes for videos based on precise description of events in the video, using video voice transcription (i.e. text version of sound). E.g. a consumer can speak to the TV “Take me to the scene in this movie where Anna meets Mary a few days after John has left the country for his second meeting with Tom to discuss his promotion”.
• Easily model complex scenarios/simulations for analysis, simply by describing them in natural language.
• Enable a new way of hiding secret information “in plain sight”, among large natural text content. This can be done by introducing words leading to the secret information among different sentences and providing appropriate questions to another party whose answers will identify the secret words in the large text. E.g. the secret message “Turn on the light if you can meet at six” can be hidden in the following example text:
  • “More and more women want to turn to crafts as a source of additional income. They do this on the premise that crafts can be created easily at home and sold at a profit. Mary had tried to make a light out of used boxes but did not know how to sell it. It will help her if she can in the future be certain about being able to sell her art. May be you can help her. It will be best if you can meet with her within the next six days, before the holidays start. It is probably easier for her to meet at your office”.
• The first words to the answers to the following questions (as an example) reveal each word of the secret message:
  • What can women do to make a better living? – Turn to crafts (yields first word “turn”)
  • Why do they want to do that? – on the premise that crafts can be created easily (yields first word “on”)
  • What did Mary try to make but did not make any money from? – light out of used boxes (yields first word “light”)
  • What will help her? – if she can be certain of making money (yields first word “if”)
  • Who should meet with Mary? – You (yields first word “you”)
  • What will make an ideal next step for you to help Mary? – can meet Mary (yields first word “can”)
  • What type of interaction is being suggested? – meet (yields first word “meet”
  • Where is the interaction being suggested? – at the office (yields first word “at”)
  • In how many days is the meeting being suggested? – Six (yields first word “six”)
• Stringing together all first words re-generates the secret phrase “Turn on the light if you can meet at six”. Note that all capabilities of InstaNLP® are needed to correctly generate the answers and reveal the secret phrase; meaning that a conventional computer , no matter how powerful, is useless in detecting the secret phrase. Conventional cryptology techniques (e.g. encryption of text and questions) can be used additionally, if desired. Other variations of bringing words from answers together to re-generate the secret phrase can be applied, e.g. using the last/middle/N’th word of the answer etc.

For more information, visit www.Instaknow.com or Email: Solutions@instaknow.com

More patents to come…Stay tuned!