ApproxiMatch applies fuzzy logic principles to create an Internet search solution that mimics the human problem-solving process: all options are considered and the best solution is found by rejecting non-relevant, redundant or contextually wrong alternatives.
Relevancy – Conventional search engines often fail to identify the most relevant results. They often match a single query term but miss the user's intent. ApproxiMatch narrows the results, including only relevant items from all the dimensions that relate to the user's query.
Completeness – Conventional search engines often miss potential results that may not match any single term of the query but are relevant to the user's intent. For example, offering 'chiropractor' when the user searches on 'back pain'. ApproxiMatch readily locates and ranks these results.
Context – Conventional search solutions rank results in relation to the exact query terms, based on a fixed set of rules. ApproxiMatch ranks the results in relation to the field of search and the site's business considerations, using ad-hoc ranking criteria.
Machine Learning – ApproxiMatch recall and precision improves continuously. Search logs and personal data are analyzed on an ongoing basis, "teaching" the search engine new relationships and contextual dimensions.
How it works
The ApproxiMatch search engine is based on fuzzy logic algorithms. Fuzzy logic is a mathematical formalism used to represent uncertainty and ambiguity. Unlike the exacting true/false of Boolean logic, fuzzy logic defines degrees of truth.
(see http://en.wikipedia.org/wiki/Fuzzy_logic)
In search of the best fuzzy matches, ApproxiMatch performs the following processes concurrently and recursively:
- > Understand the different terms of the user's query and Infer relationships between them. Similar to human reasoning, the number of potential relationships between search terms is unlimited.
- > Expand the search in run-time to examine linguistic and contextual matches in multiple dimensions, based on domain-specific assumptions.
- > Sort, rank and narrow down the results without missing the original intent of the user.
Traditional search engines are not able to perform fuzzy searches because they require complex computations that can take minutes to perform – much longer than is acceptable in commercial sites. With our proprietary fuzzy search algorithms and software architecture, ApproxiMatch accomplishes this task – in a blink of an eye.
