Mining and Summarizing Customer Reviews

Loading...

Flash Player 9 (or above) is needed to view presentations.
We have detected that you do not have it on your computer. To install it, go here.

0 comments

Post a comment

    Post a comment
    Embed Video
    Edit your comment Cancel

    2 Favorites

    Mining and Summarizing Customer Reviews - Presentation Transcript

    1. Mining and Summarizing Customer Reviews Lucas Rizoli 2007-11-20 CPSC 503
      • Hu & Liu, 2004a
      • Mining and Summarizing Customer Reviews
      • Hu & Liu, 2004b
      • Mining Opinion Features in Customer Reviews
    2. From http://www.amazon.com/Pearl-Harbor-Two-Disc-Anniversary-Commemorative/dp/B00003CXTG/
    3. From http://www.amazon.com/Pearl-Harbor-Two-Disc-Anniversary-Commemorative/dp/B00003CXTG/
    4. From http://www.amazon.com/Pearl-Harbor-Two-Disc-Anniversary-Commemorative/dp/B00003CXTG/
    5. From http://www.amazon.com/Pearl-Harbor-Two-Disc-Anniversary-Commemorative/dp/B00003CXTG/
    6. From http://www.amazon.com/Pearl-Harbor-Two-Disc-Anniversary-Commemorative/dp/B00003CXTG/ Plot Acting Special Effects Historical accuracy Sound quality …
    7. From http://www.amazon.com/Pearl-Harbor-Two-Disc-Anniversary-Commemorative/dp/B00003CXTG/ I want a summary Feature-by-feature evaluation Use all reviews
    8. Major phases
      • Finding features
      • “ Pearl Harbor ’s plot is totally crap”
      • Identifying opinions
      • “ Pearl Harbor ’s plot is totally crap ”
      • Producing a summary
      • Extractive, list-style
    9.  
    10. Finding features
      • Frequent feature phrases
      • Noun phrases of 3 words or less
      • In > 1% of reviews in corpus
      • Keep compact feature phrases
      • Phrase words < 3 words apart
      • In > 1 sentence in corpus
      • Trim redundant phrases
    11. Identifying opinions
      • Opinion sentences
      • Include a feature phrase
      • Associate nearest adjective with feature
      • “ The ‘splosions are freakin’ amazing ”
    12. Uncommon features
      • Infrequent features may be useful
      • “ Kate Beckinsale’s hair : beautiful”
      • Use adjective to identify feature
      • Add nearest noun phrase to features
      • Adds spurious features?
      • Only 15–20% of features
      • Countered by ranking by frequency
    13. Word orientation
      • Positive or negative?
      • Seed set of 30 words
      • Use syn/antonymy to tag related words
      • Use WordNet to find related words
    14. new fresh original innovative +
    15. new fresh original innovative + + + +
    16. unoriginal new fresh original innovative + + + +
    17. unoriginal new fresh original innovative + − + + +
    18. unoriginal new banal hackneyed trite fresh original innovative + − + + + − − −
    19. Sentence orientation
      • Average of opinion in sentence
      • If none, use previous sentence
      • Use nearby negations
      • Near if < 5 words away
      • Invert nearest opinion
      • Use opinion in “but” clause
      • If none, invert initial opinion
    20. Summary
      • Group sentences by feature
      • Divide by orientation
      • Rank features by frequency
      • Display some for each feature
      • Pro/Con-style, by-feature
    21. Finding features 0.72 0.80 Uncommon 0.79 0.66 Redundancy 0.66 0.67 Compactness 0.56 0.68 Frequency Precision Recall
    22. Identifying opinions
      • Word orientation
      • Recall: 0.69
      • Precision: 0.64
      • Sentence orientation
      • Accuracy: 0.84
    23. Evaluation issues
      • Implicit features
      • “ Planes were small” -> historical accuracy
      • Easy for human taggers to find
      • Story-telling
      • Adjectives in feature-free sentences
      • Orientation is subjective
      • Some evaluations hard to classify
    24. Future work
      • More opinion words
      • Include verbs and other modifiers
      • Opinion strength
      • “ It’s crummy” vs. “It’s worse than death”
      • Pronoun resolution
      • What is “it?”

    + Lucas RizoliLucas Rizoli, 3 years ago

    custom

    1291 views, 2 favs, 1 embeds more stats

    A summary of two papers by Hu and Liu, used as a st more

    More info about this document

    CC Attribution License

    Go to text version

    • Total Views 1291
      • 1290 on SlideShare
      • 1 from embeds
    • Comments 0
    • Favorites 2
    • Downloads 8
    Most viewed embeds
    • 1 views on http://www.fachak.com

    more

    All embeds
    • 1 views on http://www.fachak.com

    less

    Flagged as inappropriate Flag as inappropriate
    Flag as inappropriate

    Select your reason for flagging this presentation as inappropriate. If needed, use the feedback form to let us know more details.

    Cancel
    File a copyright complaint
    Having problems? Go to our helpdesk?

    Categories