Yext, Inc. (NYSE: YEXT), the Search Experience Cloud company, today announced “Milky Way,” the latest upgrade to the natural language processing (NLP) algorithm that powers Yext Answers, Yext’s revolutionary site search product.
Headlining this milestone update is the adoption of BERT, which stands for Bidirectional Encoder Representations from Transformers. Developed by Google, BERT is an open source machine learning framework for NLP designed to better understand user searches. By leveraging BERT within Named Entity Recognition (a process to locate and classify named entities mentioned in unstructured text into predefined categories), Yext Answers improves its ability to distinguish locations from other types of entities, including people, jobs, and events.
“When we launched Yext Answers last year, we introduced a new standard of search for websites — and since launch, we’ve been improving our algorithm constantly to understand natural language questions and help businesses respond to those questions quickly and accurately,” said Marc Ferrentino, Chief Strategy Officer at Yext. “Today, Milky Way marks our latest major upgrade to the Answers algorithm, leveraging BERT technology so that businesses can answer even more customer questions with greater precision.”
Yext’s Milky Way update includes the following features:
- Improved Named Entity Recognition: By leveraging BERT, Yext Answers can now better understand the contextual relationship between search terms. When someone uses a search term that can refer to multiple things, Answers will return a more relevant result by taking into account the correct classification, whether a location, person or product. For instance, the algorithm can now better distinguish when someone is looking for a “turkey” sandwich vs. a sandwich shop in “Turkey.”
- Improved Location Detection: The update leaves behind location biasing, which requires information like popularity and proximity to identify the location a person is searching for. Now, Yext Answers will filter through locations stored by a business in their Yext knowledge graph — a database of millions of structured facts — to surface the best match. This is especially useful if a search term could refer to multiple places, such as Paris, Texas vs. Paris, France.
- Updated Healthcare Taxonomy: More than 3,000 new healthcare-related synonyms, conditions, treatments, and procedures have been added to the algorithm’s taxonomy. Now, whether a patient is searching for ailments in layperson’s terms (“pink eye”) or a provider is searching with clinical terms (“conjunctivitis”), Yext Answers will be able to understand the context and deliver the best answer.
- Improved Stemming and Typo Tolerance: The update introduces improved stemming — NLP’s ability to recognize different forms of a root word — and refined typo tolerance to better match a search term with results that include variations of that term, such as “integrate” and “integrations”. With the added flexibility, businesses can now answer questions that would otherwise surface few or no results at all.