The Problem with Current Natural Language Processing (NLP) Technology
Current AI and NLP software does not understand the meaning of words, their concepts or relationships. Current software is based on statistical analysis of words. This approach relies on complex keyword search strings, n-grams and creating exhaustive lists of words which are used to look for similar text in the hope that the words will be accurate indicators of similarity, regardless of context or nuance.
To increase the probability of accuracy, current software leverages machine learning (ML). ML is extremely expensive, requiring long lead times and human assistance to collect datasets, label, clean and process data to build and train ML models. High levels of accuracy cannot be achieved because statistical word correlation cannot capture concepts conveyed in complex human language.
The entigenlogic® Solution
Our software, Franklin™ - unlike current technology - does not rely on statistical correlation. It understands the concepts conveyed by words, converts them into entigens®, then connects them via our patented logic rules and set-theory like math. Computers can then store knowledge as an expanding set of logically connected facts, much as humans do. Users can conceptually describe a search or upload an existing document that Franklin uses to auto-generate search concepts.
Franklin™ provides a Natural Language Understanding that no other approach can. entigenlogic® holds 11 issued patents, 21 pending patents, & 160 inventions secured.
Functional Capability – Intelligent Document Processing
Franklin™ sifts through any size document collection, quickly and accurately identifies relevant or similar documents, and reduces the risk of missing documents that don’t stand out in manual or statistically-based searches. We handle all forms of broadly used documents including social media, news articles, online form fields, blogs, contract solicitations, statement of work proposals, academic journal papers, and legal documents.
Reduces Costs and Saves Time
We eliminate long lead times, human-intensive document labeling, training-sets development, and associated costs. Franklin™ enables users to quickly define, refine, and re-run searches without the need for training sets.
Provides Superior Accuracy
We ingest unstructured text and store it in a manner that enables information retrieval based on an understanding of the concepts that words convey. Users find documents that are “conceptually similar” rather than “keyword similar”.
Characteristics
Document Formats Supported:
We support the analysis of unstructured text found in the most common file formats with extensions such as docx, pdf, txt, html, and xml.
Document Access Methods:
Users upload documents through their laptop, desktop, Smartphone, Tablet, cloud server, etc. If customized access methods are needed such as uploading documents from your document repository or a feed such as Facebook, Twitter etc., a Professional Services Agreement is available.
Delivery Format:
Franklin™ is currently hosted on the Amazon Web Services (AWS) Cloud as a subscription service. (See future capabilities below.) Whether analyzing dozens or thousands of documents, our tiered subscription levels fit your budget and business needs. Accessed through a secure connection from a standard Web-browser, Franklin™ offers single sign-on (SSO) for integration with your business.
Output of Results:
The system produces a report in a tabular format which can be customized to include additional metadata fields. Results can be exported to a CSV file for off-line analysis. Franklin’s Knowledge Discovery on an Enterprise Level provides accuracy unattainable until now.
Future Capabilities
Contact us to learn more about our exciting Franklin product plans regarding Microsoft Azure and Google Cloud deployment, Alerting, and Reporting, Filter Automation, Anomaly Detection, Temporal - Spatial Tracking, Q&A, etc.
Sample: Document-Defined Filter Report
The table below is an example of the Document Similarity Results page where the user is presented with the document similarity scores of each document against the chosen reference document along with document metadata to aid in the analysis. Additional metadata fields can be added.
Sample: Manually-Defined Filter Report
The table below is an example of the Relevance Results page where the user is presented with the relevance scores of each document against a chosen filter along with document metadata to aid in the analysis. Additional metadata fields can be added.