How To Replace The Discontinued Google Search Appliance?
Google has ended support for its Google Search Appliance (GSA) without announcing a clear successor product. Many GSA customers are scrambling to find a replacement to their enterprise search problems. Smart ones are also using this as an opportunity to upgrade.
Here is the lay of the land on options each CIO has today. See the Comparison Chart for most prominent vendors.
Natural Language Search (NLS)
NLS emulates human behavior using a branch of artificial intelligence called cognitive computing. It gives results akin to asking a person who understands all the knowledge in your enterprise. NLS does not deal with documents – it deals with knowledge. It can handle structured or unstructured information with ease, and trains for a particular context. NLS also keeps learning continuously from documents as well as user behavior.
There are two main kinds of technologies that can implement NLS:
- First uses Deep Learning or related algorithms, like IBM Watson. To train a Watson application you will have to find Subject Matter Experts that can annotated huge amounts of documents. However, nobody was fired for hiring IBM.
- Second are opportunistic technologies called Tactical Cognitive Computing. These use AI algorithms that are written specifically for the purpose. They don’t need human-annotated training data, which makes them highly cost effective. Coseer is one such example.
Semantic search technology tries to find the intent of the question and answers accordingly, similar to NLS. Semantic search still works at a document level. More importantly, users need to feed various inter-relationships to the algorithm in form of ontologies. There are many semantic search vendors, of which Microsoft FAST (part of Sharepoint) is the most prominent. It allows users to assign metatags to documents, and further build ontologies for semantic search over these metatags.
Keyword Based Search
The traditional search technology depends only on keywords. It ranks documents based on frequencies. Amazon’s ElasticSearch is a very cheap, dependable and customizable option for old-fashioned keyword based search.
This is not a no-brainer. You will be surprised how similar Google Search Appliance is to keyword search. Google’s web search uses hyperlinks and user behavior to give amazing results, but in the enterprise context, both are missing.
Open Source Software
Discontinuation of GSA may be a great option for you to regain control. Depending on your license date with Google, you may have time to build a customized solution based on open source software. It gives you the ability to control your experience, define your feature set and build step by step on a roadmap.
Just a note of caution, though. In our experience there is no good open source solution to implement advanced semantic search, or natural language search.