重口味SM

Startup led by U of T alumni uses AI to help Canadians track parliamentary proceedings

""
U of T alumnus Lyndon Chan hopes to boost political engagement with Parlawatch, an online tool that scrapes official transcripts from Question Period and uses natural language processing to generate daily summaries (photo courtesy of Lyndon Chan)

, a data analytics and consulting startup co-founded by 重口味SM alumni, has created that uses AI to boost engagement with federal politics by helping Canadians track parliamentary proceedings.

Overseen by lead data scientist Lyndon Chan, Parlawatch scrapes official transcripts from Question Period in the House of Commons and uses natural language processing to generate daily summaries. It also creates visualizations that show which topics are discussed in Parliament, and the sentiments of political parties on various issues.

The goal is to help people cut through clutter, misinformation and bias so they can make up their own minds about how Canada鈥檚 political parties are representing constituents.

鈥淲e were finding it hard to follow what鈥檚 happening in Parliament, so it made sense to do this project to summarize what鈥檚 happening so the average citizen can keep track [without] having to read an entire transcript,鈥 says Chan, who graduated with a master鈥檚 degree in electrical engineering from U of T鈥檚 Faculty of Applied Science & Engineering last year.

鈥淲e try and condense proceedings in a way that鈥檚 understandable to the average person.鈥

Parlawatch creates text summaries by grouping Question Period transcripts into sets of questions posed by opposition parties and answers provided by the government.

Chan says Alphabyte Solutions, co-founded by U of T alumni Ahmad Nameh and Adam Nameh, was keen to demonstrate the potential of natural language processing technologies. 鈥淲e saw that there was a lot of potential for practical use, but other companies either didn鈥檛 know about it or didn鈥檛 have the expertise to push it through,鈥 he says. 鈥淲e knew there were advanced algorithms out there, including for text summarization, and thought a good use case would be government data.鈥

Parlawatch鈥檚 text summaries work by grouping Question Period transcripts into sets of questions (posed by opposition parties) and answers (from the government), using AI to summarize the exchanges, and then grouping them according to party and date. The platform offers summaries for Question Period exchanges as far back as 2001. The sentiment analysis feature makes use of existing machine learning models that have been trained to assess sentiment on other data sets.

With Canadians re-electing a Liberal minority government this week, Chan provided a few examples of how citizens can use Parlawatch to glean insights into opposition parties鈥 stances toward the government on various issues, and their criticisms of the ruling party.

鈥淚f you want to see what the different parties think about the government on different topics, I recommend checking out the sentiment analysis feature,鈥 he says, referring to a tool shows the average sentiment of each opposition party鈥檚 questions directed to the government that can be filtered by topic, with a sentiment of zero being most negative and one being most positive.

鈥淔or example, you can set a filter on the topic of COVID-19 and see what the parties are thinking about the government鈥檚 handling of the pandemic over time.鈥

This tool can also show what parties are choosing as their main criticisms of the government, by allowing users to see which topics invoke the most negative questions from the opposition parties.

There鈥檚 also the topics timeline, which Chan says can showcase the main themes that are discussed in Parliament.

鈥淎nd you can check out the daily summaries because they show you 鈥 in text form 鈥 what鈥檚 actually happening.鈥

In addition to engaging regular citizens, Chan says he also envisages Parlawatch being a useful tool for journalists and news organizations down the road.

鈥淲e hope to build this product so that we can offer our services to different news organizations across Canada, so it becomes a part of their regular news arsenal,鈥 he says. 鈥淟et鈥檚 say a news outlet wants to search for a particular quote from a parliamentary session, we can go back and crop the video of that automatically, so they don鈥檛 have to do it manually.鈥

For the time being, however, Chan says Alphabyte Solutions is focused on finding ways to use AI to help regular Canadians engage with what鈥檚 happening in federal politics. 鈥淥耻谤 hope is that this can become something the average person can look to regularly as a go-to source to keep track of what鈥檚 happening in Parliament.鈥

Adam Nameh, Alphabyte鈥 s co-founder and chief data officer, says Parlawatch ultimately hopes to fill a void when it comes to using AI and technology to help Canadians engage with politics.

鈥淧arlawatch really made us rethink how the average Canadian consumes this information. Sitting through an entire parliamentary session is mostly out of the question, and most people choose to shut the blinds instead,鈥 Nameh says.

鈥淥耻谤 and other data-driven tools which they can go to. Meanwhile, our House of Commons sittings were perfectly transcribed and were an open field waiting to be harvested.鈥

In the future, Nameh says Parlawatch is also seeking to generate engagement on social media with its and TikTok videos.

 

 

 

The Bulletin Brief logo

Subscribe to The Bulletin Brief

UTC