Google’s Neural Matching Algorithm
Article: https://www.searchenginejournal.com/google-neural-matching/271125/
This article, “What is Google’s Neural Matching Algorithm” was one of the most interesting articles I have ever read! More specifically, it introduces the concept of Neural Matching Algorithms or ad hoc retrieval which “contrasts with standard information retrieval (IR) systems that rely on text-based signals in conjunction with (Page et al., 1999; Kleinberg, 1999) and/or user feedback (Joachims, 2002).” However, the article dives a bit deeper; outlining that this introduced algorithm’s use as a secondary part of the ranking process: first traditional methods of ranking are applied then the ad hoc retrieval segment is employed.
Furthermore, the ad hoc retrieval works by solely matching the content of the search query with that of a given web page because “this kind of algorithm is about ‘relevance matching’”– done by evaluating not only keywords bot also their synonyms to provide a more contextual comprehension of the web page to cultivate a more holistic matching process. However, Google is still looking to the future of continued (matching) algorithm innovation: ‘according to Google’s official announcement: ‘…we’ve now reached the point where neural networks can help us take a major leap forward from understanding words to understanding concepts. Neural embeddings, an approach developed in the field of neural networks, allow us to transform words to fuzzier representations of the underlying concepts, and then match the concepts in the query with the concepts in the document. We call this technique neural matching.’”
This article is pertinent to our current learning material because it relates to ranking systems of information on major web search platforms which corresponds to our previous lectures on Page Rank and the Search Industry. Furthermore, this article is particularly relevant to our study of web search with relation to PR update rule because the ad hoc retrieval works independently of link networks.