Why Users Search
Mostly when users search they are trying to do one of these things…
- Find a specific document
- Find a collection of documents
- Answer a question
- Find someone
- Discover something helpful
Users have an expectation of a “google” like search experience when we search using any product.
How Users Search
- The user will start with a question
- The system will then try to formulate a database query
- The user will then review the results and filter them to narrow down by specific criteria
The expectation is that users will easily be able to find exactly what they are looking for, just like they do when they do a google search.
Keyword Search Challenge
The terms users search for are not synonyms, topics or concepts and may not be in the desired documents. Users have a question and the formulation of a query but the system is not intuitive.
When document authors create documents they do not think about how someone is going to search for it or the questions that users want to answer.
Machines Interpret Search
For example when a user searches for…“Pictures of people having fun”
The user wants to search to…
- Pictures: find a picture
- People: with 1 or more people
- Fun: labelled as “fun”
What Users Want
- Intuitive experience
- Ability to ask a question
- Results that match their intent
- Relevant results
- Ability to filter those results
Solving The Search Problem
- What is the user’s intent?
- What are users searching for?
- How do they think about filtering?
- What questions do they want to answer?
When you understand user intent the answer you get is subjective so you need to constantly monitor the questions that users are trying to answer so that you can provide the best possible answers.
Try To Enrich Metadata
Building a pipeline and applying machine learning can analyse metadata and constantly apply changes to improve the search experience.
Create Rich Metadata
- Extract and enrich names, places, sentiments, etc…
- Standardise and apply industry terminology, catalogue or hierarchy
- Apply user terminology and translate technical jargon
Applying metadata to documents will improve the search experience by making it easier for users to find what they are looking for. This can be done with Machine Learning.
Machine learning is a data science technique that allows computers to use existing data to forecast future behaviours, outcomes, and trends.
By using machine learning, computers learn without being explicitly programmed.
Machine learning solutions are built iteratively, and have distinct phases:
- Preparing data
- Experimenting and training models
- Deploying trained models
- Managing deployed models
How Machine Learning Works
- Pattern Detection
- New Signal Identification
- Custom Signals Based on Query
- Image Search
- Word similarity detection
- Query Clarification
While machine learning isn’t (and probably never will be) perfect, the more humans interact with it, the more accurate and “smarter” it will get.
This could be alarming to some – bringing visions of Skynet from the “Terminator” movies – however, the actual result is likely a better experience with technology that gives us the information and services we need, when we need it.