Hello,
everyone welcome to the second webinar of the LearnAI series my name is Rodrigo
Souza I am a data scientist with the Microsoft research and AI team.
We have a special guest today Liam do you want to introduce yourself?
Yeah, thanks Rodrigo. Hi Everyone, My name is Liam Cavanagh and I am a program manager
i work on the content search and intelligence team which is part of the cognitive
search that Rodrigo is going to be talking about today let's go
today we'll talk about cognitive search and how it was use it with the jfk files
project we start with the microsoft vision and platform for ai then we
have cognitive search constants and related technologies and for
the jfk project you see architecture demo in the call
At the end we have ten minutes for a q and
a session is small disclaimer before we go any further this is not
the session about the related technologies work especially how they work together.
This is a microsoft vision for AI.
we can browse and apply end-to-end solutions from the Azure AI gallery
You can build your own solutions using the Azure platform.
You can interact with intelligence agents like Cortana or bots.
And, can use the better AI intro applications like
office365 or even windows itself?
And, this is the ai platform infrastructure services and tools.
Can use any operational system and
employ your models on frames in the cloud or even to IoT device
for the jfk project we use it Pre-built AI AI ON DATA
AI Computes the similarities in a few minutes?
After some research i chose these definitions so we can better understand what cognitive search is.
Artificial intelligence computing which
simulate human-human perception.
Cognitive form from a to perspective process often
intellectual activity search from a web perspective
experience of a natural low friction way to interact with applications
and data finally the definition for cognitive search extract relevant
information from big and diverse data sets in users context.
Perception drives to context and we can
say cognitive search is cognitive skills with a search engine.
If I'm going too fast please let me know.
And, what are the microsoft related technologies for
cognitive search. first one we see is cognitive services
the value of ai isnt about fancy algorithms but how to make it
easy to use.
As you can see we have api to simulate human
cognitive skills. This service are prebuilt in preach changing deep learning
models polished on Azure to accelerate your AI project.
They have free cheers.
can be used with just a few lines of code and work across platforms
like ios andriod and windows.
But, the jfk project we use a computer vision and entity linking api.
In the cognitive service portal we have
a directory for apis grouped by category.
in the quality of service for row if you use a few clicks
you can see pricing and documentation of all the fbi you need.
Example the computer vision api use in the jfk project
is free up to five thousand transactions per month is
works we've curl or any other language capable to call a recipe
api also you can download all documentation pdf format
It also has demos
allow you to see expected jason results for each fbi
sample major is provided but you can upload your own you do these
in a few minutes.
with computer vision api.
In the jfk project you read
text from images in pdf files if the entity linking
api you get the wikipedia id of the terms
and expressions extracted from the computer vision processing I just image so
the two technologies work together these two api work together one after the other.
Azure search the second microsoft technology use it for cognitive search energy
in the JFK project.
It's a cloud search path service for web and
mobile applications.
It's easy to scale up and down and has ninety nine-point nine per cent
SLA. It creates an index if metadata about your data it's not
data integration it's not data replication its an index
with information about your data?
Provides natural language processing and
have special features for same structure or unstructured data?
Which features are these, spelling mistakes?
Geo spatial data and not only plotting but also distance intersections
so on suggestions ranking paging
Highlighting and facets.
However, working with relational database you can imagine how complicated it
is to do all of these we've sequel lots of IFs.
Like percent queries very bad for performance.
Sub queries.
Cursors
Secondary indexes so it's pretty hard to index
unstructured data we have sequel work with structured data
that's where Azure search helps the most and how it works?
It has corey profiles so can define ranking in relevance for each term
this
process has a default algorithm but you can create your own.
Azure search provides process.
Processing capabilities for multiple language removing stop wards or
provide a concept understanding of the data you can use Lucien microsoft algorithms for
this possible data sources are Azure sequel end string column.
CosmosDb the files on Blob storage include including office files
pdf csv xml and so on and also
Azure tables are key values storage on Azure.
Không có nhận xét nào:
Đăng nhận xét