Anand Ranganathan, Chief AI Officer at Unscrambl
Conversational AI is getting more and more widely used for customer support and employee support use-cases. In this session, I'm going to talk about how it can be extended for data analysis and data science use-cases ... i.e., how users can interact with a bot to ask analytical questions on data in relational databases.
This allows users to explore complex datasets using a combination of text and voice questions, in natural language, and then get back results in a combination of natural language and visualizations. Furthermore, it allows collaborative exploration of data by a group of users in a channel in platforms like Microsoft Teams, Slack or Google Chat.
For example, a group of users in a channel can ask questions to a bot in plain English like ""How many cases of Covid were there in the last 2 months by state and gender"" or ""Why did the number of deaths from Covid increase in May 2022"", and jointly look at the results that come back. This facilitates data awareness, data-driven collaboration and joint decision making among teams in enterprises and outside.
In this talk, I'll describe how we can bring together various features including natural-language understanding, NL-to-SQL translation, dialog management, data story-telling, semantic modeling of data and augmented analytics to facilitate collaborate exploration of data using conversational AI.
Best VIP Call Girls Noida Sector 39 Call Me: 8448380779
Data Con LA 2022 - Collaborative Data Exploration using Conversational AI
1. Collaborative Data Exploration using
Conversational AI
Anand Ranganathan
Co-Founder & Chief AI Officer
anand@unscrambl.com
2. How do people consume data & analytics today?
– through charts & dashboards
3. What’s wrong with dashboards
Limited or no drill-downs
Don’t know how the data for the dashboards
is produced
Can’t ask a slightly different question
Representative of an opinion; easy to
cherry-pick stats
Can be misleading
4. Analytics & BI tools have some ways to go
In making organizations truly data-driven
“Where can I find my
data? I don’t know
which database, table,
query or tool to use”
“The presentation
is tomorrow and
the BI team is
busy. What
should I do?”
“Why are
there so many
dashboards?”
26.5%
of organizations report
being data driven in 2022,
down from 37% in 2017.
5.
6. Allow any user to ask text or voice questions of their data
and
receive back a natural language + visual analysis
of statistically relevant and actionable insights for that user.
What is Conversational Analytics?
*Note that in this talk, we focus on structured data stored in
relational format (e.g. SQL databases, Excel sheets, etc)
7. Qbo: Natural language conversations with data
within collaboration platforms
Hey QBO,
how many
policies are
expiring in
Oct 2022?
Hey QBO,
why were
new
acquisitions
in Feb lower?
Why have
sales
dropped in
the North
region this
month?
AI-Powered,
Data Analyst
11. A (very simplified) overview of NLU pipeline
number of trips in winter 2017 by age and gender
Entity Recognition & Construction
SELECT anon_1."age group", anon_1.gender, count(*) AS "Count"
FROM (SELECT "TRIP_ANALYSIS".end_station_id AS "end station id", "TRIP_ANALYSIS".program_id AS "program id", "TRIP_ANALYSIS".start_station_id AS "start station id",
"TRIP_ANALYSIS".bikeid AS bikeid, CASE WHEN (:birth_year_1 - "TRIP_ANALYSIS".birth_year < :param_1) THEN :param_2 ELSE CASE WHEN (:birth_year_2 -
"TRIP_ANALYSIS".birth_year < :param_3) THEN :param_4 ELSE …, "TRIP_ANALYSIS".gender AS gender
FROM "TRIP_ANALYSIS"
WHERE ….
Identification of Query type and mapping to known concepts in DB
Type: Aggregation Query on Trips table with a group-by and a filter; age -> derived from birth year attribute; gender -> gender attribute; in winter
2017 -> 2017-12-23 and 2018-03-19 (filter)
Generate DB-specific SQL query
Get results, decide on visualization and narratives, and present back to user
13. ● Users don’t know what to ask
● Users don’t know how to ask
● Users may pose questions in
an ambiguous manner
● Users may use terms not in the
dataset
Key Challenge : Bridging the gap between users and
data
14. ● Users don’t know what to ask
● Users don’t know how to ask
● Users may pose questions in
an ambiguous manner
● Users may use terms not in the
dataset
Key Challenge : Bridging the gap between users and
data
● Data may be modeled in a variety of
ways
● Hidden semantics and assumptions
behind different tables and columns
● Data may be incomplete, unclean
● Data may be spread across silos
15. Common scenario we encounter in our
deployments :
Qbo doesn’t understand the user
17. Collaborative data exploration to the rescue
● Many organizations and teams contain people with different levels
of data skills
● In the past, the data-have-nots were dependent on the data-haves
● Conversational AI can help them collaborate. Allow the
data-have-nots serve themselves sometimes, and the data-haves
jump in when they can’t
18. Direct
Connection
to data
Natural
Language
Queries
Collaborate with internal and external users to explore
& share data, insights and reports
With the right access control restrictions in place
Internal
Business Users
3rd Party
Partners
Collaborate with
Partners in the
context of “Teams”
or “Channels”
19. Interesting challenges around collaborative
data access
● How should access control work?
○ Based on the user asking the question?
○ Based on the user in the group channel with the “least” privilege?
● How do we keep track of context around multiple parallel conversations
in a group chat
● How do we keep track of versions and group edits in a collaborative
dashboard?
○ Especially when the charts are interactive or can be refreshed