Parlawatch.ai
Parlawatch is an AI powered analysis tool for Canada’s daily Question Period proceedings. It assists in keeping average citizens and political analysts informed as well as the government accountable.
The tool summarizes and predicts sentiments for Question Period proceedings in the Canadian House of Commons, dating back to 2001.
Intuitive and user-friendly, Parlawatch utilizes machine learning for pre-trained sentiment analysis models in order to produce insightful predictions.
Summarizing and analyzing text transcripts, Parlawatch monitors Parliament and generates concise reports in order to inform curious citizens, journalists, and political analysts alike.
Topic Timeline:
View topics by weekly frequency of mentions throughout a given Parliamentary session.
Daily Report:
Summarizes the entire day’s Question Period proceedings as a back-and-forth conversation between the parties and the government.
Question/Answer Awards:
Ranks the highest-scoring quotes from Question Period daily for sentiment: Irony, anger, sadness, and optimism.
Sentiment by Topic:
Parlawatch scores the average sentiment of the various parties toward the government in any given topic.
Technologies used
- Azure
- Microsoft SQL Server
- Python
- Random Forest
- Serenity
- .NET Core
- TypeScript
- HTML
- CSS
How Parlawatch.ai Works
Training
We train our AI by first loading and warehousing historic House of Commons Question Period data.
Custom ground-truth summaries are then annotated and used to train machine learning/natural language processing models.
Prediction
Parlawatch generates predictions by loading and warehousing the latest House of Commons Question Period digital Hansard data and applying it to previously trained text summarization and sentiment analysis models.
Sentiment results are graphed and can be viewed through a variety of filters, allowing the user to hone in on the information that they are searching for.
Furthermore, quotes with the highest-scoring sentiments will then have the corresponding video snippet extracted and automatically downloaded. These snippets are then uploaded onto Twitter using a bot.