Kalicube Tuesdays Special Edition: A Deep Dive into KELM and MUM.

Language models combined with Knowledge Graphs are the future of search (aka a deep dive into KELM and MUM).

Payoff: the coming years will see a radical transformation in Search guided by Knowledge Graphs and the increased abilities of language models. MUM and KELM are just getting started, and if you want to prepare, ahead of this seismic shift for SEO, join this webinar on June 29th.
-----

00:00 Dawn Anderson, Bill Slawski, Andrea Volpini and Jason Barnard
03:09 MUM - a multitask recommender system
04:31 How search engines collect information for the knowledge panel
06:06 The evolution of indexing; moving towards a language model giving conversational experience
07:41 Dawn explains what a language model is
09:28 Unstructured data: Wikipedia and T5 training models
11:46 Google’s teachable moments
14:30 Degrees of trustworthiness of web sources
15:41 Cost optimization strategies in the multimodal search engine
17:25 How ranking is changing via changes in indexing
19:21 Language agnostic Search engines - is that something realistic?
20:45 The dangers of relying on data from several different sources
22:08 Linguistic crossover between languages
25:23 KELM defined
29:40 Evaluating the toxicity and accuracy of the language model dataset
31:02 Does search engine optimization still make sense in the future with KELM and MUM?
33:21 Google is becoming a multimedia engine
37:18 Will SEOs have more work in the future?
39:56 Making the search intent more granular for ambiguous and broad terms
47:44 What SERP to expect for ambiguous queries
49:09 Dawn’s piece of advice to websites with cannibalisation problems
50:09 How do we optimize our content for language models?

-----------
Bill Slawski, Dawn Anderson, Andrea Volpini and Jason Barnard discuss the ins and outs of MUM and KELM and how they might affect the evolution of the Knowledge Graph.

More specifically, we'll look at the relationship between language models, Google algorithms and the Knowledge Graph and shed light on things that we might usefully do to improve our entities in the Knowledge Graph using language (so, how language models and knowledge graphs interact).

Bonuses (Boni ???!!):
Jason will share the results of some of the 500+ experiments optimising entities for the Knowledge Graph.

Andrea promises to reveal if there is a way to do SEO for Machine Learning language models like GPT-3!

BIll will share his insights into the patents that underpin these technologies.

Dawn will dispel some of the myths and tell us where Google is 'at', where it is going and how fast this is likely to move forwards in the coming months and years.

Organised by Kalicube in partnership with Wordlift.

Part of the Kalicube Tuesdays series.

[BrianSchalkxSEO]

What a powerhouse of guests. Great explanations. Especially the T5 architecture fron Dawn