November 2022 Summaries

Neural search technology is transforming the customer experience by delivering relevant content that corresponds with the underlying meaning of a query, rather than just matching words. This approach reimagines the search experience, whether users are searching for information on massive public websites or looking for obscure data points in enterprise databases. Neural search recognizes context and understands the distinctions between categories of related products, providing a better overall product discovery experience. It also enables chatbots that consider the context when surfacing solutions, creating a new class of conversational agents that derive meaningful value from complex web support chat and call data. Additionally, neural search can help customers go a step further toward self-service, directing them to the exact location in content that answers their questions, reducing the burden of customer service and enabling proactive analytics for decision-making.

Recent advancements in natural language processing (NLP) using neural networks and large language models (LLMs) are transforming the traditional search experience by enabling organizations to provide more relevant and instantly available results, thereby enhancing the customer journey and opening new business opportunities. These neural search systems address the limitations of keyword-based searches by understanding concepts and context, delivering accurate results even with misspellings or acronyms, and improving product discovery on eCommerce platforms. This innovative technology extends to chatbots, creating more effective conversational agents that enhance customer support by considering context and providing solutions based on past interactions. Furthermore, neural search facilitates proactive self-service, allowing users to independently resolve queries and access information efficiently, thus reducing the need for support interactions and offering business insights through self-service analytics. As search leaders like Google and Amazon invest in these capabilities, turnkey solutions like Vectara enable smaller organizations to integrate advanced NLP content discovery into their operations, enhancing customer engagement and loyalty with instant, personalized responses.

Using metadata fields within Vectara can enhance search capabilities by providing context to indexed content. There are several techniques for using metadata, including query filtering, boosting, and results enrichment. Query filtering allows users to narrow down search results based on specific criteria, such as geographic area or timestamp. Boosting increases the confidence score of a result based on attributes associated with the text. Results enrichment displays additional information with search results to make them more engaging. Vectara provides three categories of metadata fields: automatic metadata, custom metadata, and pass-through metadata. Users can specify filterable metadata fields when creating a corpus and indexing content, and then use these fields in query filtering and boosting. By effectively using metadata fields, users can improve the relevance and engagement of their search results.

Large language models (LLMs) like Vectara are transforming search by providing highly relevant results without prior language configuration, and their effectiveness can be enhanced through the use of metadata fields. Metadata, which provides context to data, enables more precise search results by allowing users to filter content based on various criteria, such as geographic location or presence of personally identifiable information (PII). This approach supports privacy by tagging content with metadata indicating PII presence, thereby enabling filtered results based on user permissions. Additionally, metadata can be used to boost or bury search results and enrich them by attaching related information at index time, improving the user experience by offering more contextual insights and relevant links. Vectara's search capabilities leverage these metadata techniques, employing automatic, custom, and pass-through metadata fields to enhance query filtering, result boosting, and enrichment, all while supporting cross-language searches and providing users with natural language answers.