[{"data":1,"prerenderedAt":159},["ShallowReactive",2],{"project-30955699":3},{"id":4,"name":5,"line":6,"line_source":7,"title":8,"url":9,"domain":10,"maker":11,"posted_at":12,"year":13,"points":14,"comments":15,"rank_in_year":16,"shown_in_year":17,"top":18,"tier":19,"category":20,"subcategory":21,"format":22,"keywords":23,"reception":28,"sentiment":29,"stance_mix":30,"scored_comments":34,"aspects":35,"strip":38,"post_text":39,"what":40,"said":71,"maker_in_thread":101,"shape":107,"grid":109,"attempts":120,"same_week":121,"maker_shows":31,"maker_hidden":152,"reader_version":153,"card":154},"30955699","Programmatic","a REPL for creating labeled data","title","Show HN: Programmatic – a REPL for creating labeled data","https:\u002F\u002Fprogrammatic.humanloop.com","programmatic.humanloop.com","jordn",1649417731,2022,26,5,1333,2570,52,"shown","ai","data labeling","library",[24,25,26,27],"weak supervision","labeled data","nlp annotation","labeling functions","quiet",null,{"positive":31,"neutral":32,"mixed":33,"critical":33},2,1,0,3,{"execution":36,"idea":37},{"positive":32,"critical":33},{"positive":32,"critical":33},[34,15,33,15,33],"Hey HN, I’m Jordan cofounder of Humanloop (YC S20) and I’m excited to show you Programmatic — an annotation tool for building large labeled datasets for NLP without manual annotation.\n\nProgrammatic is like a REPL for data annotation. You:\n\n\n```\n  1. Write simple rules\u002Ffunctions that can approximately label the data\n  2. Get near-instant feedback across your entire corpus\n  3. Iterate and improve your rules\n\n```\n\nFinally, it uses a Bayesian label model [1] to convert these noisy annotations into a single, large, clean dataset, which you can then use for training machine learning models. You can",{"problem":41,"audience":47,"pricing":51,"motivation":55,"time_to_build":29,"stack":59,"openness":68,"stage":69,"team_size":70},{"value":42,"source":43,"evidence":44},"Getting labeled data for NLP is a blocker; hand annotation is slow","maker",[45],{"id":46},"post",{"value":48,"source":43,"evidence":49},"Software engineers and data scientists building NLP datasets",[50],{"id":46},{"value":52,"source":43,"evidence":53},"Free package",[54],{"id":46},{"value":56,"source":43,"evidence":57},"Labeled data is a blocker for businesses using NLP; wanted a faster way to build datasets",[58],{"id":46},[60],{"value":61,"source":62,"evidence":63},"Skweak (HMM label model)","commenter",[64],{"id":65,"by":66,"text":67},"30956624","razcle","Hi Raza here, one of the other co-founders. I know that HN likes to nerd out over technical details so thought I’d share a bit more on how we aggregate the noisy labels to clean…","free","beta","company",{"praise":72,"objections":87,"requests":88,"questions":89,"alternatives":96},[73,83],{"value":74,"source":62,"evidence":75},"Speeds up domain expert labeling and model development cycles",[76,80],{"id":77,"by":78,"text":79},"30955841","hmaguire","I've been using this for the past month or so in Beta and it's fantastic. I'm a DS at an NLP startup and it's totally changed the way we develop new tagging and classification…",{"id":81,"by":78,"text":82},"30956944","It allows us to leverage domain expertise much more quickly and efficiently. Since we we can extrapolate from just a handful of examples, we can \"solve\" the easy parts of the…",{"value":84,"source":43,"evidence":85},"Interactive feedback makes labeling feel game-like",[86],{"id":46},[],[],[90],{"value":91,"source":43,"evidence":92},"What has it changed about your process?",[93],{"id":94,"by":11,"text":95},"30956592","Ace! That's awesome to hear. What's it changed about your process?",[97],{"value":98,"source":43,"evidence":99},"Snorkel and skweak weak supervision packages",[100],{"id":46},{"comments":31,"answered":33,"key":102},[103,104],{"id":94,"text":95},{"id":105,"text":106},"30955787","Just like to clarify that this goes beyond a rule-based system. Rules can get you pretty far[1] but this improves on that by intelligently discounting the bad rules using weak…",{"commenters":34,"max_depth":34,"maker_replies":31,"first_comment_mins":108,"lifespan_hours":34},15,[110,112,114,116,118],[65,33,34,111,66,33],"maker\u002Fneutral\u002Fexecution",[77,32,15,113,78,33],"project\u002Fpositive\u002Fexecution",[94,32,33,115,11,32],"maker\u002Fneutral\u002Fnone",[81,32,15,117,78,33],"project\u002Fpositive\u002Fidea",[105,31,33,119,11,32],"maker\u002Fneutral\u002Fidea",[],[122,127,132,137,142,147],{"id":123,"name":124,"points":125,"category":126},"30946792","Redditle.com",448,"web_saas",{"id":128,"name":129,"points":130,"category":131},"30970911","Explordle",390,"learning_science",{"id":133,"name":134,"points":135,"category":136},"30956775","Eight Colors",368,"games",{"id":138,"name":139,"points":140,"category":141},"31002171","indieblog.page",241,"productivity",{"id":143,"name":144,"points":145,"category":146},"30992587","Multiplayer",221,"infrastructure",{"id":148,"name":149,"points":150,"category":151},"30991235","howto-wasm-minimal",217,"dev_tools",false,"v2",{"id":155,"name":5,"line":6,"ai":152,"cat":20,"year":13,"points":14,"comments":15,"rec":28,"s":29,"top":18,"kw":156,"maker":11,"domain":10,"tier":19,"attempts":33,"strip":157,"url":9,"mix":158,"praise":74,"objection":-1,"sub":21},30955699,[24,25,26,27],"35050",[31,32,33,33],1793991044046]