A free fortnightly newsletter on Taxonomy, Thesauri & Ontology and Semantic Publishing

Training machine learning to improve enterprise search

Training machine learning to improve enterprise search

September 20, 2018

What users expect from search has evolved. In fact, you can demarcate the evolution as B. G. and A. G.—before Google and after Google. Before Google, users were happy with the results that had a semblance of relevance. After its advent, users expect “psychic search” even from enterprise search engines. Psychic search is a type of search that, like Google, just seems to know the intended meaning of a query. According to Miles Kehoe, founder of New Idea Engineering, Google and other public web search and ecommerce sites have evolved along with the changing user expectations. He adds that they have morphed from “web search sites” into “machine-learning matching systems”.

When you search on any of the large web properties—Google, Yahoo, MSN, Amazon and such—your search provides clues, which are known as signals. When these signals are processed by machine learning (ML) or artificial intelligence (AI) tools and the results are integrated with the search, they automatically influence the relevance and the displayed results.

However, ML needs to be set up correctly to enable the process of training an instance with search. Furthermore, the training will be used over time to incorporate user behavior, which will then provide more signals. As ML/AI tools generate profiles for both users and content, the increase in the number of signals translates into “people like you” search results that most users will find helpful. This is particularly true when it comes to enterprise search.

Enterprises do not have the expanse of content or the number of users that are available in the large web search and shopping sites. Therefore, they might have fewer signals, but the signals they have are more likely to be accurate. This means that the tools and data employed in an enterprise are more precise. Additionally, the tools that work splendidly on large websites will likely deliver great results even on smaller sets of content typically found in enterprises. Similar to how ML is trained for tuning by large internet behemoths, the training and tuning of ML/AI tools in enterprises should be an ongoing process.

Click here to continue reading on why training and tuning of ML/AI tools are important.

Brought to you by Scope e-Knowledge Center, an SPi Global Company, a trusted global partner for Digital Content Transformation Solutions, Knowledge Modeling (Taxonomies, Thesauri and Ontologies), Abstracting & Indexing (A&I), Metadata Enrichment and Entity Extraction.

Please give your feedback on this article or share a similar story for publishing by clicking here.

Comments are closed.

Start typing and press Enter to search