In the AI Startup of the Week, the editorial staff of ai.nl is featuring promising AI startups, their innovations, solutions and challenges. In this twelfth episode, we are taking a look at Zeta Alpha, a Dutch AI startup building the next generation enterprise search and insights platform.
We have repeatedly seen through our coverage of tech companies, AI fueled companies, and even small and medium enterprises, that the challenge facing them often is organising the data or knowledge necessary for their data science and AI teams to draw powerful insights. Amsterdam-based Zeta Alpha solves this very pain point.
Formally known as Zeta Alpha Vector BV, this artificial intelligence (AI) startup was founded by Jakub Zavrel on May 1, 2019. The startup, according to Crunchbase, has raised around €1M in seed funding. In its three years of existence, the startup has been unique for its 100 per cent focus on AI and data science. Here is a look at how Zeta Alpha is building a smarter way to discover and organise knowledge for AI and data science teams.
Discovering and Organising data remains a challenge
To really understand what Zeta Alpha does, we need to first understand how humans behave when they want to learn something in this modern world. Whenever faced with a question or need to learn something, people often turn to a search engine platform like Google. With 90 per cent market share, Google is essentially the window to world’s knowledge for a lot of people.
However, if you are a company with lots of data, you can’t just Google your way to access all that information or knowledge. Zeta Alpha is building the enterprise-grade search engine that will help companies to streamline their data and knowledge infrastructure, which is necessary for AI and data science teams.
It is important to understand at this juncture that every company is essentially becoming an AI company now. To become an AI fueled company, there is a need for AI and data science teams, who have the primary job of analysing data to deliver insights that are critical to the success of the business. While big companies like Google, Microsoft, and Facebook can do this discovery process in-house, there are a number of companies who struggle with data.
In the Netherlands alone, a survey about small and medium enterprises has shown that these companies want to become AI-driven but often struggle with understanding where to start. For such companies, it may become easy to set up an AI team but streamlining the data pipeline will continue to be a challenge. With that in mind, let us look at how Zeta Alpha solves this problem with its product designed with AI and data science teams in mind.
Neural network to aid discovery of knowledge
At the core of what Zeta Alpha is building is deep learning based neural network models, which have seen great advancements thanks to very large language models based in the transformer architecture. Zeta Alpha says this has allowed for “complete rethinking of the basis of document understanding and search.”
With its neural retrieval and discovery technology, Zeta Alpha is able to overcome the lexical gap, also referred to as the keyword gap. In a traditional system, a query is matched by referring the search phrase (or keyword) with keywords added by the authors in their documents. If they don’t match then the search won’t render accurate results. However, Zeta Alpha is able to train its neural network model on very large document collections that understand “prior semantic relationships between words and phrases.”
The neural network also solves the problem faced by traditional systems to understand word order, relationships, and context. With a neural network trained on a large language model, Zeta Alpha says it is able to deliver search results that preserve the meaning implied by the structure of the sentence. A keyword-based search system will never be able to deliver such a result.
Search is just one part of the data equation but the real ability of Zeta Alpha’s tech is to aid discovery. On its website, the Dutch startup makes it relatively easy to understand. For starters, the discovery of existing and new knowledge is provided through the Search API, wherein it offers both keyword-based search as well as the aforementioned neural vector search. This search works both at the document, passage and sentence level, and naturally relies on a number of filters and ranking functions.
The discovery part comes in via the ability to look beyond internal data sources and check external data sources such as “social media mentions, author influence, or code popularity.” The easiest way to understand Zeta Alpha’s discovery system would be to look at Google again. Whenever you search for something on Google, the search giant offers a combined search results first and granular search options like Shopping, News, etc atop.
Zeta Alpha has a feature called ‘Find Similar’ that allows the user to query by example from documents or collections using other search operators and filters. Further, it also has system modules called visual analytics, two-dimensional maps to get a quick overview of a large body of documents. Expert search is a module that allows users to connect with internal or external experts which sounds a lot like LinkedIn.
What about services?
All that technology built by Zeta Alpha eventually comes down to exit as a service. This service is mainly aimed at companies that have a large amount of unstructured data and are not satisfied with classic keyword search. The first service offered by Zeta Alpha is called Enterprise Search, which is a code neural discovery platform tailored for different customers.
The Enterprise Search helps with indexing of all enterprise content using state-of-the-art neural search with customised processing workflows. It can also be designed for finer access controls, SSO and API integration. There is also the option to configure system components such as crawlers, connectors to third-party systems, or document processing steps to new schema and knowledge domains.
All of this would be meaningless without a data analytics service and Zeta Alpha does that by offering detailed reports analysing the AI and data science literature. The real highlight of Zeta Alpha’s product is not essentially its ability to make knowledge discovery easier but also customise it based on field and individual needs.
What to read next?
- ⚙️ How Dutch startup Researchable partners companies to drive innovation through data
- 🍉 How Dutch startup Watermelon is making it easier for businesses to automate customer service with AI chatbot
- 🖇️ Here’s how Dutch startup Supplai is the linking pin between AI and business with its automation product