Viet Yen Nguyen, Andréa B. Duque, Jules Belveze, Lukas D. Baker, Astrid Harsaae, Pedro Pessoa, Inês Oliveira
{"title":"Atlastic Reputation AI: Four Years of Advancing and Applying a SOTA NLP Classifier","authors":"Viet Yen Nguyen, Andréa B. Duque, Jules Belveze, Lukas D. Baker, Astrid Harsaae, Pedro Pessoa, Inês Oliveira","doi":"10.1109/DSAA53316.2021.9564190","DOIUrl":null,"url":null,"abstract":"We report on the main challenges we had to overcome while bootstrapping, commercializing and scaling our automated sentiment analysis solution in the media intelligence industry. Over the span of four years, our solution evolved from a run-of-the-mill component to a bespoke in-house bred flagship feature that, to our knowledge, trumps those of our industry peers. From all the learnings to date, the most significant one has been shifting our data labeling efforts from a third party to a completely in-house controlled process — a move that brought us noticeable improvements in inferencing quality.","PeriodicalId":129612,"journal":{"name":"2021 IEEE 8th International Conference on Data Science and Advanced Analytics (DSAA)","volume":"193 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2021-10-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2021 IEEE 8th International Conference on Data Science and Advanced Analytics (DSAA)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/DSAA53316.2021.9564190","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0
Abstract
We report on the main challenges we had to overcome while bootstrapping, commercializing and scaling our automated sentiment analysis solution in the media intelligence industry. Over the span of four years, our solution evolved from a run-of-the-mill component to a bespoke in-house bred flagship feature that, to our knowledge, trumps those of our industry peers. From all the learnings to date, the most significant one has been shifting our data labeling efforts from a third party to a completely in-house controlled process — a move that brought us noticeable improvements in inferencing quality.