jurafsky&martin_3rdEd_17 (1).pdf

Since wordnet has these relations it is often used

Info icon This preview shows pages 332–334. Sign up to view the full content.

Since WordNet has these relations, it is often used ( Kim and Hovy 2004 , Hu and Liu 2004b ). After a seed lexicon is built, each lexicon is updated as follows, possibly iterated. Lex + : Add synonyms of positive words ( well ) and antonyms (like fine ) of negative words Lex - : Add synonyms of negative words ( awful ) and antonyms ( like evil ) of posi- tive words An extension of this algorithm has been applied to assign polarity to WordNet senses, called SentiWordNet (Baccianella et al., 2010) . Fig. 18.6 shows some ex- SentiWordNet amples. In this algorithm, polarity is assigned to entire synsets rather than words. A pos- itive lexicon is built from all the synsets associated with 7 positive words, and a neg- ative lexicon from synsets associated with 7 negative words. Both are expanded by drawing in synsets related by WordNet relations like antonymy or see-also. A clas- sifier is then trained from this data to take a WordNet gloss and decide if the sense being defined is positive, negative or neutral. A further step (involving a random- walk algorithm) assigns a score to each WordNet synset for its degree of positivity, negativity, and neutrality. In summary, we’ve seen three distinct ways to use semisupervised learning to induce a sentiment lexicon. All begin with a seed set of positive and negative words, as small as 2 words (Turney, 2002) or as large as a thousand (Hatzivassiloglou and
Image of page 332

Info icon This preview has intentionally blurred sections. Sign up to view the full version.

18.3 S UPERVISED LEARNING OF WORD SENTIMENT 333 Synset Pos Neg Obj good#6 ‘agreeable or pleasing’ 1 0 0 respectable#2 honorable#4 good#4 estimable#2 ‘deserving of esteem’ 0.75 0 0.25 estimable#3 computable#1 ‘may be computed or estimated’ 0 0 1 sting#1 burn#4 bite#2 ‘cause a sharp or stinging pain’ 0 0.875 .125 acute#6 ‘of critical importance and consequence’ 0.625 0.125 .250 acute#4 ‘of an angle; less than 90 degrees’ 0 0 1 acute#1 ‘having or experiencing a rapid onset and short but severe course’ 0 0.5 0.5 Figure 18.6 Examples from SentiWordNet 3.0 (Baccianella et al., 2010) . Note the differences between senses of homonymous words: estimable#3 is purely objective, while estimable#2 is positive; acute can be positive ( acute#6 ), negative ( acute#1 ), or neutral ( acute #4 ) . McKeown, 1997) . More words of similar polarity are then added, using pattern- based methods, PMI-weighted document co-occurrence, or WordNet synonyms and antonyms. Classifiers can also be used to combine various cues to the polarity of new words, by training on the seed training sets, or early iterations. 18.3 Supervised learning of word sentiment The previous section showed semi-supervised ways to learn sentiment when there is no supervision signal, by expanding a hand-built seed set using cues to polarity similarity. An alternative to semi-supervision is to do supervised learning, making direct use of a powerful source of supervision for word sentiment: on-line reviews .
Image of page 333
Image of page 334
This is the end of the preview. Sign up to access the rest of the document.

{[ snackBarMessage ]}

What students are saying

  • Left Quote Icon

    As a current student on this bumpy collegiate pathway, I stumbled upon Course Hero, where I can find study resources for nearly all my courses, get online help from tutors 24/7, and even share my old projects, papers, and lecture notes with other students.

    Student Picture

    Kiran Temple University Fox School of Business ‘17, Course Hero Intern

  • Left Quote Icon

    I cannot even describe how much Course Hero helped me this summer. It’s truly become something I can always rely on and help me. In the end, I was not only able to survive summer classes, but I was able to thrive thanks to Course Hero.

    Student Picture

    Dana University of Pennsylvania ‘17, Course Hero Intern

  • Left Quote Icon

    The ability to access any university’s resources through Course Hero proved invaluable in my case. I was behind on Tulane coursework and actually used UCLA’s materials to help me move forward and get everything together on time.

    Student Picture

    Jill Tulane University ‘16, Course Hero Intern