The examples that I chose for my benchmark demonstrate that Lexiathan maintains accuracy and performance even on severely degraded input. On less corrupted input, Lexiathan runs significantly faster and is even more accurate.
Lexiathan also doesn't have any edit distance parameters that need to be configured, so there is no "tuning" required. In particular, it's worth mentioning that using a very large dictionary, e.g. 500,000 words, often degrades accuracy rather than improves it, and likely increases memory usage and latency as well.
Regarding Norvig's 98.9% figure--this seems to be from Norvig's own made-up data. In the real world, users often generate misspellings that exceed 2 edit distances in many use cases (OCR, non-native speakers, medical/technical terminology, etc), and published text (often already spell-checked) doesn't reflect the same level of errors. And in any case, Norvig's spell-checker apparently only achieves an accuracy of 67% on its own chosen benchmarks, so clearly the 98.9% figure is not a realistic reflection of actual spell-checker performance, even for an edit distance of 2. Lexiathan is extremely accurate and retains high performance even on heavily degraded input, and the benchmark data (and demo) that I presented reflect that.
I reported the results faithfully and I believe these results reflect the performance that users would typically see running SymSpell in the browser, using the default configuration. If I had increased the edit distance, then the latency performance gap between Lexiathan and SymSpell would have been even larger, and then arguably I would have been gaming my metrics by not using SymSpell as it is configured.
Regarding dictionary size: The dictionary size (as you can verify from the gist) was 82k words. I didn't specify the dictionary size I used for Lexiathan, but it was 106k words.
Lastly, three of the words in the benchmark have edit distances greater than three:
For the curious: The article doesn't name the ally, but is probably referring to Qatar, which hosts the Al-Udeid Air Base, where the USAF has a significant presence. Geographically, the US could potentially strike Iran from Qatar; they are only separated by the Persian Gulf.
It probably did damage the runway a bit, but given that these ancient planes are no longer in production, the cost to repair whatever damage was done to the plane (looks pretty extensive, from the video) is likely much more expensive than whatever work is needed to fix the runway.
In defense of the students: the types of films that you watch as part of a film study curriculum are generally not the same as what most cinema-goers are watching. For example, "Man with a Movie Camera"--or 150 minutes of someone's black and white movie about the life of urban pigeons... present-day film students who grew up watching movies with tight editing, fast cuts, high resolution color and sound, and quick narrative payoffs are not going to respond to these movies the same way that people did a century ago.
This is not to say that historical films lack value; but sitting all the way through them with rapt attention is not necessarily as easy as you'd imagine.
I thought about doing this sort of thing a while back by parsing code into an AST + using a debugger API. Using an LLM is an interesting approach, but I would be concerned whether (a) it is reliable and (b) whether it is deterministic and replicable.
Lexiathan also doesn't have any edit distance parameters that need to be configured, so there is no "tuning" required. In particular, it's worth mentioning that using a very large dictionary, e.g. 500,000 words, often degrades accuracy rather than improves it, and likely increases memory usage and latency as well.
Regarding Norvig's 98.9% figure--this seems to be from Norvig's own made-up data. In the real world, users often generate misspellings that exceed 2 edit distances in many use cases (OCR, non-native speakers, medical/technical terminology, etc), and published text (often already spell-checked) doesn't reflect the same level of errors. And in any case, Norvig's spell-checker apparently only achieves an accuracy of 67% on its own chosen benchmarks, so clearly the 98.9% figure is not a realistic reflection of actual spell-checker performance, even for an edit distance of 2. Lexiathan is extremely accurate and retains high performance even on heavily degraded input, and the benchmark data (and demo) that I presented reflect that.