Under the Hood

The Problem

 

We've talked to hundreds of patent practitioners who all have the same problems with search technology: 

  1. Developing a good search query.

  2. Identifying relevant results.

Dorothy solves both problems using natural language processing and machine learning.  Here's how. 

The Search

 

Screen Shot 2019-01-09 at 3.33.13 PM.png

The end of keyword searching!

No keyword query necessary.  Simply insert a description of the searched subject matter in the large text box, and let Dorothy do the work.   

Untitled design (14).png
 

Search what you mean NOT what you say…

Prior art is filled with jargon.  How do you know you've incorporated the right synonyms to capture the all important references?  Dorothy identifies terms in your query that are part of our proprietary thesauruses and replaces them with a "token." The token replaces your word with every synonym in our thesaurus.  Our thesaurus was created by indexing the entire patent database, so your days of worrying that you missed an important synonym are over.      

 
Untitled design (14) (1).png
 

Get Specific…

We both know most of your query describes things in the prior art.   Identify what's new in the "Novel Feature" box and Dorothy will move references describing similar elements to the top of the list! 

 
Untitled design (14) (1).png

Processing

 

The patent database in virtual space.

Word embeddings are vector representations of particular words that  are capable of capturing context of a word in a document, semantic and syntactic similarity, relation with other words, etc. 

Word embeddings allow Dorothy to plot the words of entire patent database in virtual space.  Terms having used in similar context cluster together and generally describe similar technology. 

Word embeddings for the terms in your query are overlayed onto this plot.  Neighboring word embeddings represent words from the data base that describe similar technology.  

 
Words related to kitchen appliances (yellow) in this plot are clustered in one area this graph and words related to tools (red) are clustered in another area.

Words related to kitchen appliances (yellow) in this plot are clustered in one area this graph and words related to tools (red) are clustered in another area.

Untitled design (14).png

The Results

 

 
Dorothy identifies relevant text! Highlighted snippets of text in the identified references describe elements of your search query. No need to search through the entire disclosure to find relevant text.

Dorothy identifies relevant text! Highlighted snippets of text in the identified references describe elements of your search query. No need to search through the entire disclosure to find relevant text.

You Read My Mind

Dorothy's results are sorted  by relevance with the most relevant publications at the top.  You won't get far down the list before you find what you're looking for.  But, Dorothy is still learning, so we built in re-ranking, which re-sorts results based on references you view and references you remove from the list.  The more you search the better your results get without revising your query.   

Untitled design (14) (1).png