Livio Robaldo, Luigi Di Caro, Alessio Antonini Users of - - PowerPoint PPT Presentation

livio robaldo luigi di caro alessio antonini
SMART_READER_LITE
LIVE PREVIEW

Livio Robaldo, Luigi Di Caro, Alessio Antonini Users of - - PowerPoint PPT Presentation

Livio Robaldo, Luigi Di Caro, Alessio Antonini Users of the web post on social network, forum, blogs, etc. opinions about commercial products. That huge amount of data is highly precious for marketing


slide-1
SLIDE 1

Livio Robaldo, Luigi Di Caro, Alessio Antonini

slide-2
SLIDE 2

… … …

  • Users of the web post on social network, forum, blogs, etc.
  • pinions about commercial products.
  • That huge amount of data is highly precious for marketing

researches.

  • But, since it is very huge, we need automatic methods to analyze it.

… … … … … … …

slide-3
SLIDE 3

Positive/negative opinions are only a subsets of the relevant

  • comments. It would be worth annotating also suggestions,

comparisons, questions or simply observations. Moreover, they do not correspond 1:1 with sentences: sometimes they are found in modifiers and are intertwined with other comments.

slide-4
SLIDE 4

[Robaldo and Di Caro, 2013] L. Robaldo and L. Di Caro: OpinionMining- ML, Computer Standards & Interfaces. Vol. 35, n.5, 2013. Proposes a new XML formalism for tagging comments strongly facet-

  • riented and grounded on syntactic representation.

It allows to tag both appraisals and suggestions, observations, comparisons, question, and any other kind of speech act. It allows to describe and organize facets into a domain ontology, that may be designed off-line. Afterwards, the concepts in the

  • ntology are related to the speech acts that talks about them.

It allows to deal with complex linguistic constructions, e.g. coordinations, as the structure of the XML tend to mirror the

  • ne of the (dependency) syntactic representation.
slide-5
SLIDE 5

Crawler

Txt files in utf-8

Parser Tule

Dependency Trees Domain ontology

XmlBuilder

OpinionMining-ML documents