Introduction to Pair Search
Last updated
Last updated
Pair Search is developed as part of OGP’s .
Pair Search is a search engine that aims to enhance your search experience by using state-of-the-art technology to deliver more accurate results.
With advanced keyword matching and semantic search, it ensures you find the relevant information you need easily. The intelligent snippets feature also helps by presenting concise and relevant snippets, putting valuable information right at your fingertips.
Pair Search currently supports these datasets:
Hansard Parliamentary Reports: All publicly available data on Hansard from 1955
Supreme Courts Judgments (Judiciary): All written judgments issued by the Supreme Court of Singapore since 2000
Legislation: Includes all Current Acts and Subsidiary Legislation
Do note that the search results of each dataset is independent from each other. When querying, you will receive results from only one dataset at a time. However, you can easily toggle between the two datasets to access the relevant information.
If there are any other datasets that you would like to see supported, please let us know through our .
Select the dataset that you want to query.
Type in your search query using keywords, similar to how you would on Google.
See relevant results with snippets to quickly gauge the usefulness of the results.
Supported Search Operators
Pair Search supports search operators similar to Google. Some common examples include:
Using quotation marks to search for an exact phrase
Using the minus sign to exclude specific words from the search results
💡 Tip: If you’re looking for specific speakers on Hansard, you can use quotation marks in your search. E.g. “Edwin Tong” Singapore Sports Hub concert strategy
When you perform a search, the results you see are sorted by relevance by default.
How It Works: Our hybrid ranking algorithm evaluates documents using a combination of keyword-based and semantic-based metrics:
Keyword-Based Metrics: These traditional metrics match your search terms with those found in the documents.
Semantic-Based Metrics: Utilising the latest advancements in machine learning-based retrieval, such as ColBERTv2, these metrics understand the context and meaning behind your search terms, providing a more nuanced and accurate ranking of the results.
Sorting Options: If you prefer, you can change the sorting order of your search results. Here are the available options:
Relevance (Default): This option uses our hybrid ranking algorithm to bring the most relevant documents to the top of your search results.
Date: This option sorts the search results by their publication or modification date, with the most recent documents appearing first.
Filtering Options
Our datasets come equipped with specialised filters tailored to each type of data.
For the Hansard dataset: Dates, speakers, report types, parliament, session, and sitting.
For the Judiciary dataset: Dates, catchwords, and court type.
For the Legislation dataset: Date commenced, last amended, and legislation type.
These tailored filters ensure that you can efficiently navigate and pinpoint the exact information you need from each dataset.
Designed to help you quickly navigate large volumes of information, our AI-powered analysis tools provide specific functionalities to meet your needs. They offer a preliminary overview, allowing you to explore the landscape of a topic, cover all bases, and discover perspectives you might not have considered before diving deeper into research with the provided citation links.
Hansard Timeline and Speaker Analysis
Gain a comprehensive understanding of parliamentary proceedings with our Hansard analysis tool. It provides:
Timeline Analysis: Analyse discussions and debates over time to identify trends and significant events.
Speaker Analysis: Analyse contributions from different speakers to understand perspectives, frequencies, and key points.
Judiciary Case Summaries and Key Principles Overview
Case Summaries: Quickly grasp the essence of judicial decisions with concise summaries of key cases.
Overview of key principles: Identify and extract fundamental key principles, aiding in research and reference.
Legislation Summaries and Key Areas
Key Areas Regulated: Highlight the main areas covered in the legislation.
Summarisation: Summarise each piece of legislation for quick understanding.
Drafting Assistance and Result Categorisation (Available for all datasets)
Enhance your document preparation and organisation with our drafting assistance and categorisation tools:
Drafting Assistance: Write custom prompts to synthesise your search results and tailor Pair Search’s response to your needs.
Sample prompts:
"Summarise the key arguments for and against the proposed environmental bill based on the search results.”
"Identify the key recommendations from the recent committee report on education policy."
"Prepare an answer for a question regarding the government's stance on renewable energy investments.”
Result Categorisation: Automatically categorise search results into relevant topics and themes, streamlining your review process and making it easier to find the information you need.
Use our Intelligent Search Suggestion feature to expand and refine your searches, ensuring you don't miss anything of interest.
Disclaimer: The feature makes use of an AI language model and might not be exhaustive. Any content it generates may contain errors, inconsistencies, or outdated information, and does not represent official government views.
Please fact-check and verify all AI-generated content before use, and ensure that you are in compliance with any applicable laws, regulations, instruction manuals, or practice directions currently in force.
Pair Search uses a multi-phased ranking approach that gradually filters and ranks the retrieved documents. Less computationally expensive ranking features are utilised for the earlier phases, with the more accurate but expensive measures only being utilised in the latter phases. This keeps results accurate, while ensuring that the searches are completed with minimal latency.
The final ranking is based on a hybrid score that integrates semantic, keyword-based, and ColbertV2 scores. Additionally, a freshness score is incorporated into the retrieval step, to favour more recent reports.
For troubleshooting and support, please contact us through our and let us know the issue that you’re experiencing.
If you have any feedback, please let us know via our . We welcome your feedback to improve your search experience with Pair Search.
