rosette Sign Up Information

Dear Isabella,

What qualifies as "smart" technology? or "AI"? These are questions we take very seriously here at Basis Technology. We even wrote a book about it.

For us, "smart" often means adding intelligence at every stage of a natural language processing system. It's a search algorithm that knows "Bob" often means "Robert" and that "Telegraph" is very similar to "Telephone." It's a language identifier that chooses a different model depending on how much text you have to work with and detects different languages within a document. 

We recently took up the problem of address matching. While many solutions geo-code, normalize, or perform a lookup, we felt that a truly "smart" address matcher should look at the fields of text, the way a human does, spotting obvious errors that classic approaches might miss.

We’ve written more about our smart address matching approach in this blog post. And we are also sharing two stories about due diligence and social media monitoring applications that use Rosette to make their applications smarter.
 

---

Smart Address Matching

Rosette now offers smart matching for names, addresses, and dates

Astaire Avenue, Garland Drive, Lamarr Avenue, Skelton Circle, and Hepburn Circle are real street names in Culver City, CA, and equally prone to spelling errors as person names. In fact, many components of an address are essentially names: streets, cities, states/provinces, or buildings. Rosette has been widely adopted for its intelligent fuzzy matching of person, location, and organizational names, but up until now did not specifically handle these “embedded names” in addresses.

Rosette now applies algorithmic smarts to postal addresses the same way it does for person names in the LABS (beta) release of the new /address-similarity  endpoint.

Check it out!

 

---

More Relevant Search for Due Diligence

Red Flag Group saves time for clients by implementing Rosette intelligent name search

The Red Flag Group’s global due diligence services cover 194 countries and 54 languages. “The biggest frustration of our customers is sifting through all the irrelevant results [false positives]...Clients want to be notified when ‘Acme Brick’ in Russia was in the news for presenting some sort of client risk, but not for ‘Acme Bricklaying’ in France... The biggest dilemma is clients can’t not look at all the results.” 

Learn how Rosette alleviated these headaches with intelligent name matching.

Read More

 

---

Media Monitoring at Global Scale

Sensika’s 360 intelligence monitors 10x the media with 1/10th the human labor

Sensika is a “media seismometer” for its clients, aiming to detect the subtle “tremors” that most media monitoring tools might miss. Whether it’s knowing a company is in trouble two weeks before it files for bankruptcy, or surfacing pricing complaints two hours after a product launch, clients rely on Sensika to find relevant media mentions in the markets that matter to their business.

“Big companies navigate like submarines. Using sonar, they send out a signal and listen for the feedback,” Konstantin Christoff, CEO of Sensika, said. “Correct data that comes late is useless for them. They need correct, useful, and timely data.”

Read More

 

---

Happy Coding!

The Rosette API Team
 

Dear Isabella,

One of the things that make Rosette better than other common text analytics alternatives is the quality of our training data. For our development teams, this means a lot of time spent on data annotation. Not just annotating, but also assessing the quality of the annotation. 

In the past, we gave our annotators common open source tools to do their work. However the workflow was clunky, and the model couldn't be evaluated until a large number of documents had been tagged. So recently, we developed a proprietary tool, Rosette Adaptation Studio.

Faster Annotation with Rosette Adaptation Studio

Rosette Adaptation Studio uses natural language processing and active learning to speed up the annotation process and deliver: 

• AI-assisted tagging - Base linguistics segments documents into useful chunks; entity extraction informs higher order annotations; and semantic similarity measures the uniqueness of each document.
• Efficient annotation - Instead of tagging more-of-the-same documents, the algorithm selects documents for annotation which are more likely to increase the model's coverage and accuracy.
• Iterative model evaluation - The model is continually retrained and evaluated as newly tagged documents are added, so you can stop tagging when the model reaches your target accuracy. The result is annotators tag fewer documents than before to achieve accurate models.

Rosette Adaptation Studio is currently in beta. We and our partners have had great success using it.  Check out this demo and let us know if you have a big annotation project in the works that could use a boost!



Want to learn more about building custom NLP using Rosette? Watch our recent product update webinar.

You can also learn more about data annotation by reading our recent blog series on evaluating NLP.

Happy Tagging!

The Rosette API Team

Dear Isabella,

Often entity extraction is thought of as a generic extraction of people, places, and organizations from documents for metadata creation. However, what if you want to extract entities that are very specific to your business? This could be employees, customers, healthcare providers, schools, vendors, bike store owners or any other database you keep. 

We now offer a fast way to link entity mentions in documents to the entries in your database using the new Custom Database Connector in Rosette Entity Extractor and Linker. Rosette can extract your entities from unstructured text and link them to their corresponding IDs in your database. This intelligent matching looks for all known aliases and disambiguates using context surrounding the entity in the document and database. Every match comes with a dynamic linking score, with which you can automatically insert high-scoring documents to your database or forward borderline matches to human review.