Article: When smart algorithms beat artificial intelligence -brute force

[Gurigraphics]

It can be useful for those looking to improve transcription quality
Mainly for voice assistant executing "known commands", such as do this or that
And even more so when memory and processing resources are scarce: embedded, IoT, etc

https://dev.to/gurigraphics/when-smart-algorithms-beat-ai-brute-force-57mj

[liam-mceneaney]

Good article, the wchess example actually already uses levenshtein distance!

whisper.cpp/examples/wchess/libwchess/WChess.cpp

Line 74 in 8f253ef

const float sim = similarity(txt, k_prompt);

whisper.cpp/examples/common.cpp

Line 786 in 8f253ef

float similarity(const std::string & s0, const std::string & s1) {

[Gurigraphics]

Thanks for the complement
Devs who are not from this area generally don't even know that it exists or what it is for
In the article I also mentioned others: Jaro–Winkler Similarity, Jaccard Similarity

Another possibility is to use "embeddings" for semantic similarity
In this case the model would be able to recognize that "Brazil" is a "Country" and "orange" is a "fruit" or "color"