Home

Welcome to the Auto-Extractor wiki! Here you will find information related to Auto Extractor

---

The current status

• Clustering the web pages based on style and structure
• Scalable on Apache Spark

Roadmap

• Auto extraction of content
• Port to Map Reduce and thus plug into Apache Tika

---

Quick Start

Cluster a bunch of htmls on file system

Not documented yet. Look for FileCluster in the code

Cluster Nutch Output Segments on Spark

This functionality is provided by autoextractor-spark module.

• build the autoextractor-spark module : mvn clean package
• run : java -jar target/autoextrator-spark-*.jar

 Usage:
 -list VAL            : List of Nutch Segment(s) Part(s)
 -master VAL          : Spark master url (default: local[2])
 -sw (--sim-weight) N : weight used for aggregating structural and style
                        similarity measures.
                        Range : [0.0, 1.0] inclusive
                        Notes :
                                0.0 disables structural similarity and only style
                        similarity will be used (it is faster)
                                1.0 disables style similarity and thus only structural
                        similarity will be used
                         (default: 0.0)
 -workdir VAL         : Work directory.

• put segment content part paths to a file. For example list.txt contains :

/ext/tg-ext/sites/old/batch1/batch1/segments/20151003201017/content/part-00000/data
/ext/tg-ext/sites/old/batch1/batch1/segments/20151003201017/content/part-00001/data
/ext/tg-ext/sites/old/batch1/batch1/segments/20151003221050/content/part-00000/data

• Run the job : java -jar autoextractor-spark/target/autoextractor-spark-0.1-SNAPSHOT.jar -list list.txt -workdir out-4 -master local[4]