$ links -dump https://news.ycombinator.com/item?id=17205865 | grep -Eio '(php|python|react|javascript|java|objective-c|objective|object-c|object|haskell|rails|m

$ links -dump https://news.ycombinator.com/item?id=17205865 | grep -Eio '(php|python|react|javascript|java|objective-c|objective|object-c|object|haskell|rails|mongo|lamp|docker|aws|devops|android|ios|django|ruby|rails|react|swift|node|docker)\b' | tr A-Z a-z| sort| uniq -c | sort -rn 87 react 63 python 42 javascript 39 aws 33 devops 29 rails 28 ruby 28 node 23 android 22 java 19 ios 19 docker 15 django 12 haskell 8 php 7 swift 5 mongo 2 object 1 objective-c 1 objective Used links to get just the text and not the HTML markup. Used grep -O to emit just the part of the regex which matches so we can feed it all into a pipe that emits a single histogram. By ordering the pattern we can match "javascript" before "java" and not need to add the space or "," after "java" to exclude javascript (you could also use "\b" with grep -E).Case-insensitve and case-folded. Added a few other keywords. Since grep only emits the first match on a given line, this won't count properly if multiple keywords appear on one line. Why did you search for "mean"? Edit: here's another approach that splits the output into words first so that we can count multiple keywords on one line, along with a couple tweaks to the pattern to eliminate some false positives: $ links -dump https://news.ycombinator.com/item?id=17205865 | tr -s '[[:punct:][:space:]]' '\n' | tr A-Z a-z | grep -E '\b(php|python|react|java|object|haskell|rails|mongo|lamp|docker|aws|devops|android|ios\b|django|ruby|rails|react|swift|node|docker)' | sort | uniq -c | sort -rn 87 react 63 python 42 javascript 38 aws 33 devops 29 rails 28 ruby 27 node 23 android 21 java 19 docker 18 ios 15 django 12 haskell 8 php 7 swift 6 nodejs 5 reactjs 5 mongodb 5 mongo 2 objective 2 object 1 reactnative 1 reactive Another issue: this only grabs the first page of comments.