The created software platform scans social media and other Internet sources to create reports on the reputation of its customers - among which are the European Commission, Air France and other large customers. Like most of this software, the application is engaged in the analysis of semantics, linguistics and heuristics. However, like any other system with machine data analysis, their tool often has problems with such subtle parts of human speech as sarcasm and irony - and, it seems, just Spotter managed to overcome this problem - let their leaders admit that the result so far is far from ideal, and itβs still too early to fully trust the car. The percentage of recognition reaches 80%, and, according to the authors, a few years ago even a similar result was unthinkable - then sarcasm was identified in 50% of cases. The authors say that the algorithm works with 29 languages ββ(including Russian and Chinese), and most often they have to deal with the recognition of messages about poor service levels.Source: https://habr.com/ru/post/185772/
All Articles