Sim Gamboa

🔆 @seraconlp - Search and Assistant AI

[Sim Gamboa @pinoystartup]

Published a NeurIPS 2019 trip report.
Hope you’ll enjoy it and find it useful.
https://t.co/yaHn8tzvKz

[Satya Nadella @satyanadella]

At the end of the day, the greatest privilege of my job is working with people who are driven by mission. These last 5 days, I saw people across OpenAI remaining calm and resolute in driving their mission despite all that was happening around them. And I saw people across Microsoft remain focused on our mission and serving our customers and partners, stepping up to help in every way. This is what I’m especially thankful for going into the Thanksgiving holiday. Thank you for your resolve and for the work you do each day to advance AI safely and responsibly and distribute its benefits to all of humanity.

[Alex Rampell @arampell]

@Suhail There are only three jobs:
-build product
-sell product
-help the people building or selling the product

[Massachusetts Institute of Technology (MIT) @MIT]

We hope you enjoyed the 21st day of the month, in this 21st week of the 21st year of the 21st century.

[Sim Gamboa @pinoystartup]

Wanted to form an opinion of Viome’s claim of its approach to biological age clock - one that utilizes microbiome metatranscriptonic and blood transcriptonic methods. Current standard is Dr. Steve Horvath’s DNA methylation measurements. Viome sells personalized food supplements.

[Sim Gamboa @pinoystartup]

Studying Viome’s Biological Age calculation model: “An accurate aging clock developed from the largest dataset of microbial and human gene expression reveals molecular mechanisms of aging” https://t.co/pMjckCYg7G

[Sim Gamboa @pinoystartup]

-Pres. Joe Biden

[Sim Gamboa @pinoystartup]

https://t.co/h6rYXn41M4

[Sim Gamboa @pinoystartup]

@TheBencuya @paulg My own tally from analyzing data trickles

[Sim Gamboa @pinoystartup]

@paulg

[Sim Gamboa @pinoystartup]

"Representation" and "matching" are some other aspects.

Also, as these data points are usually "sparse" to "signal" itself completely or enough, we "borrow" representation data from other models that are trained in richer corpus.

[Sim Gamboa @pinoystartup]

When the task then is "relevance ranking", some of the feature categories one needs are:

* query context
* item context
* user context
* user-query context
* user-item context
* environment context
* business context
* and maybe query-item context

[Sim Gamboa @pinoystartup]

Coarser "filters" are used in the different search domains. "Intentions" however I would argue are more "granular" sub-categories of filters that may not be binary/explicit in form.

[Sim Gamboa @pinoystartup]

In session searching, one can also hypothesis the importance of the "sequence of intentions" to achieve an ultimate goal. Modeling this may also aid in helping users reach their session goal faster.

[Sim Gamboa @pinoystartup]

Search is "business specific". It can be abstracted though as there are N categories of corpus sources. Search intention can be any of N x M. M being the intentions in each N. User at one point may intend 1 or a few intentions. Being able to know it helps in ranking and diversity

[Sim Gamboa @pinoystartup]

In search, a "click model" can be one of the features in Learning to Rank. Research has found out that there are user search "intent biases" accompanying clicks. If one can remove it, relevance ranking can be more effective.