Summer of Math Exposition

Presented by 3Blue1Brown 3blue1brown

How similar are two words? | Needleman–Wunsch

Audience:

Tags: computer-sciencelinguisticsoptimization

We present a method to calculate the similarity between words based on their phonetic transcription (their pronunciation) using the Needleman–Wunsch algorithm. It was originally devised to compare amino acid sequences of two proteins. Here, we explain every detail of the algorithm, including a Rust implementation. The video is accompanied by a paper, see the GitHub repo for more details: https://github.com/Splines/phonetics-graph


Analytics

6.84 Overall score*
33 Rank
14 Votes
9 Comments

Comments

5

This video explains the application of a mathematical algorithm to a linguistic topic in a visually appealing and easy to follow way, but lacks clear motivation about why the topic is important other than to demonstrate the algorithm.

7.7

Great video! Clear animation and presentation, good pacing, good explanations.

I’m conflicted on the focus on French rather than English because on one hand I totally understand if you’re more familiar with the former, and I even appreciate going outside the usual examples. On the other hand, it does limit the reach to those who are more familiar/willing to be unfamiliar with French. Not a huge deal, but something to consider if audience/reach is a concern.

Only real thing I would probably recommend is to try and vary the narration style a bit. It’s a bit monotone at times, some variation in the tone of voice can help keep the energy going.

5

Nice way to introduce DP and make it intuitive. It was that it promised to run the 600k^2 on the dictionary but left it for another video but valid clickbait ig.

1

It is good to understand how phonetic sounds could be compared with a solution that includes linear algebra. I would like to see a future video on that graph

8

Explaining the algorithm by working through a specific example made the explanation very understandable

8.2

The video is absolutely amazing. Completely easy to follow and beautiful. I think the most helpful part could be the notion of “distance” between words and the reasoning about the “most optimal path”. So interesting.

As feedback, I would maybe put more examples with the matrix to see the general method and give some examples and applications of how It could be useful (maybe in AI voice systems). Also the more general problem of information theory, error correction using distances etc.

So good, I loved It. Thanks.

7.1

Pacing, production quality, are all phenomenal! Topic was presented very clearly, and the choice of topic is interesting.

4.6

I felt like this was half a video. What’s the purpose of this? Who cares? What happens with different chosen values?

5.5

This video is quite well animated, and the animations are very useful for helping understand what’s going on. Although I enjoyed the video, I don’t think the creator really captured my interest with why we were interested in calculating these similarity scores. Yes, I can imagine some reasons, but I think it would be better if the viewer didn’t have to imagine the reasons. Along the same lines, it wasn’t fully clear to me why the pseudo-code portion of the video was included. Again, I can imagine some good reasons, but the bottom line is that the video was not as compelling as it might have been . Yes, the visuals were very nice, but if it’s not clear why we are going down a particular path, then the viewer’s interest is likely to wane.