Information Extraction from Tweets

Competence Center:  Information Retrieval and Machine Learning
ContactProf. Dr.-Ing. Sahin Albayrak, Sascha Narr
Partner: SearchMetrics 


Social networks and services like Twitter offer a wealth of information as thousands of users publicly exchange information. Among the messages are often comments about brands, products and persons. An automated analysis of the exchanged text messages allows gaining an overview of reactions and opinions about popular items. This helps identify trends and trendsetters and can offer aid for marketing decisions.

Language Identification and Semantic Analysis

In the IET project, methods are developed to first recognize which language a tweet is written in. We then analyze if the tweet contains known entities, like persons or brands. After determining if the tweet contains subjective text passages, we analyze whether subjective tweets contain positive or negative sentiments. The methods employed are adapted specifically for the use on micro text messages like tweets. They are therefore robust in dealing with very short text pieces and misspellings and can handle colloquial texts and incomplete sentences.