MetisBanner.png

BING Conversational Search

 

GOAL

Integrate best of Bing Bot Framework learnings to make Bing more efficient for conversational search scenarios. Allow users to accomplish tasks more efficiently by guiding them through their search process.  

 

SOLUTION

We identified that the conversational model encourages users to clarify and refine their queries when search engine fails to understand them. Also we can guide the users to evolve their intent to reach the right answer. Search paradigm opportunities:

  • Answer complicated questions that normally require clarifying dialog
  • Resolve user’s unspoken subject
  • Better complete user’s unspoken task

Metis (Project name) drives conversation through replying to queries with natural language questions and dialog to help guide users find what they’re looking for in a more personable way. 

 

APPROACH  

Our earlier efforts of integrating bots on SERP were great technology concepts but had low overall engagement. The bots worked well for few scenarios and we wanted to carry forward few of the aspects of bots conversation and embed them into SERP for better user interaction. We decided to leave behind the bot issues like: user value prop confusion, discoverability, chit chat model & real estate constraints. And carry forward other top approaches from bot like:

  • Guided Search: Evolve the search box to support multi-turn while maintaining what users love about traditional search
  • Query Clarification: Ask users for more information to give them the best result and boost answers accordingly
  • Quick Replies: Save users time by anticipating query refinement

 

LOW-FI / MID-FI IDEATION PROCESS  

We looked into multiple ideas of integrating clarification dialog and intent refinements across the SERP. Can share on request.

 

FINAL MODULE ANATOMY AND INTERACTION

SAMPLE FLOW

UPDATED BREADCRUMB UI & PROTOTYPE OF MODULE IN ACTION

Metis on Desktop

Metis on Mobile.