Algorithm recognizes online sarcasm

Jerusalem - a computer scientist at the Hebrew University of Jerusalem
have developed an algorithm that can detect the sarcasm in online
postings.

Using such a sarcasm detector might summaries of the researchers, or
their ranking customer reviews on websites such as Amazon are
improved. The SASI (Semi-supervised Algorithm for Sarcasm
Identification) actually works, has shown the team led by computer
science professor Ari Rappoport considering the online merchant. The
software could detect in an experiment with 77-percent accuracy
whether they are dealing in Amazon reviews with sarcastic sentences.
Given the often subtle nature of the sarcasm that is a remarkable
value.

Sarcasm on Amazon In the experiment were evaluated all records from a
total of 66,000 English Amazon customer reviews for 120 products from
SASI. To enable this, the Israeli researchers have relied on a
five-part scale for increasing sarcasm, and the values ​​one and two
are normal formulations. Initially selected sample sets of people were
classified on this scale.

Such a training set was created for the algorithm, extracted from the
word sequence patterns were to serve as a sarcastic note. So then the
66,000 reviews SASI has been studied in sarcastic terms.

One could just sarcasm detection on sites like Amazon have practical
benefits. /

To assess how accurate is the place to detect, then 15 human subjects
from a selection of 90 records have to assess the SASI has classified
as normal or sarcastic. With this basis of comparison, the computer
scientists came to the conclusion that their algorithm is correct in
more than three quarters of the cases. Personalization advantage

A sarcasm detection could Rappoport and his staff, according to sites
like Amazon just have practical benefits. Because many users would
find sarcastic reviews unhelpful. On the other hand, there would be
users who appreciate this style. Therefore, a sarcasm-detection could
be just for a personalized presentation of customer reviews of
advantage. On how easily could be the method by moving to other
languages ​​than English, the computer scientists, however, have made
any claims. The work, titled "ICWSM - A Great Catchy Name:
Semi-Supervised Recognition of Sarcastic Sentences in Product Reviews"
will be next week as part of the 4th International AAAI Conference on
Weblogs and Social Media presented next. Since the Amazon experiment,
the researchers have tested their sarcasm detection on Twitter.