Resolution extraction involves running the summarizer module to find the most frequently used sentences for possible resolutions. This approach involves sampling a few cases and manually extracting all the resolutions. It keeps only a few sentences from an email or text with valuable information, acting as a summarizer. Due to the high volume of key words and sentence structures, other evaluation metrics for machine translation and other text generation tasks in NLP, like bilingual evaluation understudy (Bleu) scores are also used. Using additional metrics reduces the total volume of sentences extracted from the logs after the subsequent iterations.
