Neural Search for next-gen machine language understanding

Enterprise-scale Search Cloud platforms are all about the inclusiveness of results and relevance. However, almost all search technologies today still use variations of two algorithms that have been around as early as the 1970s. While there have been many updates and versions of these themes, improvements in search are hard. They require specialized skills to constantly tune and customize search results, which is too costly for most enterprises.

As a result, on average, individual employees spend 400 hours each year searching for information to do their job, and often do not find what they need, when they need it. This can result in millions of dollars in operational costs and lost business opportunities. Relevance has been a challenge at scale — most employees expect a Google-like experience at work, searching seamlessly across all the content created within their organization's firewall. Unlike web search, enterprise search is far more complex. The information is contained in multiple applications, building silos of knowledge, and the volume of content is enormous. A large organization can store billions of documents, which all need to be searched securely.

Moreover, most enterprise search systems produce relevant search results using statistical or world-based methods but are less effective regarding natural language queries. To overcome those limitations, we have developed neural or language-based search to materially improve the relevance of search results for natural language queries by using deep neural nets to foster information retrieval.