Improving Recall of Regular Expressions for Information Extraction

Learning or writing regular expressions to identify instances of a specific<br/>concept within text documents with a high precision and recall is challenging.<br/>It is relatively easy to improve the precision of an initial regular expression<br/>by identifying false positives covered and tweaking the expression to avoid the<br/>false positives. However, modifying the expression to improve recall is difficult<br/>since false negatives can only be identified by manually analyzing all documents,<br/>in the absence of any tools to identify the missing instances. We focus on partially<br/>automating the discovery of missing instances by soliciting minimal user<br/>feedback. We present a technique to identify good generalizations of a regular<br/>expression that have improved recall while retaining high precision. We empirically<br/>demonstrate the effectiveness of the proposed technique as compared to<br/>existing methods and show results for a variety of tasks such as identification of<br/>dates, phone numbers, product names, and course numbers on real world datasets