Ashequl Qadir University of Wolverhampton, UK ashequl.qadir@wlv.ac.uk
Outline Introduction Related Works Methodology Results Improvement Challenges Conclusion and Future works
Introduction
Review Examples
Introduction Customers write reviews concerning their satisfaction and criticisms . Products have their functional features . (Ex. Performance, Portability etc . ) Customers study reviews of product features prior to buying . Reviews contain: Opinions ( Ex. It is very easy and simple to use ) Factual Information ( Ex. It has a pink metal case ) Opinions can be: General (Ex. I am very happy with this product ) Product feature specific (Ex. It works nicely – Performance feature)
Introduction Explicit product feature : “The design looks nice ” – Explicitly mentioned design feature. Implicit Product feature: “ I found this product really useful for transport ” – Implicitly present portability feature. Web contains unstructured opinion texts. Summarizing feature specific opinions by reading is time consuming . Opinion Sources: Review sites (epinions.com), e-commerce sites (amazon.com), blogs etc. Goals: Opinion Detection Product feature assignment
Related works Opinion identification: • Specific orientation of POS tags (Turney, 2002) • Seed list of lexicons (Godbole et al., 2007; Kim et al., 2005 ) • N-gram subjectivity clue (Riloff et al.,2003) • Word co-locations (Wiebe et al., 2001) • Dependency relations of words (Wilson et al., 2004; Fei et al.,2006) Explicit product feature identification: • By extracting noun phrases (Yi et al., 2003,2005; Hu et al., 2004) • By parts and properties of a product (Popescu et al., 2005 ) • By approaching as a classification problem (Ghani et al., 2006) • By using frequent word associations (Qadir, 2009) • By using dependency grammar graph (Zhung et al., 2006) Approach in this paper differs by: Utilizing typed dependency relations to detect opinion Using statistical methods for identifying associated words. Utilizing tf.idf weight score to assign a product feature to a review line. Not discriminating between explicit and implicit product features.
Typed Dependency Relations Words in a sentence have certain grammatical dependency relations. easy xcomp nsubj advmod cop set It incredibly was conj cc aux prt use and up to Example: It was incredibly easy to set up and use. o nsubj (nominal subject): Syntactic subject of a clause o cop (copula): Complement of a copular verb and the copular verb o advmod (adverbial modifier): Modifying adverb o xcomp (open clausal complement): Clausal complement without its own subject o aux (auxiliary): non-main verb of the clause o prt (phrasal verb particle): phrasal verb o cc (coordination): element of a conjunct and the coordinating conjunction o conj (conjunct): connected by a coordinating conjunction
Typed Dependency Relations Features of Stanford typed dependencies: More fine-grained distinctions are offered (de Marneffe, 2008) than PARC representations. (King, 2003) Ex. Breaking down adjuct into amod , xcomp , prep_of (de Marneffe, 2008) Uses Penn Treebank part-of-speech tags. Defines 55 binary grammatical relations. Relation consists of a governor and a dependent.
Typed Dependency Relations Example: It was incredibly easy to set up and use . Words representing functional feature of product: set up , use (Usability) Word representing opinion: easy Word representing degree of subjectivity: incredibly Dependency relations: o xcomp(easy, set) o conj (set, use) o prt(set, up) o advmod(easy, incredibly) o POS that can represent a functional feature: VB o Ex. uses, works, looks, fits, costs o POS that can represent a feature: Adj/NN o Ex. reliable, portable (Adj) o Ex. design, price (NN) Representations that can establish a meaningful relation between these parts-of-speech are chosen
Selected Typed Dependencies acomp - Adjectival Complement Description: An adjectival complement (acomp) of a VP is an adjectival phrase. The adjectival phrase functions as the complement (like an object of the verb) Example: Dependency Component Indication Tag POS Relation Example VBD Verb, past tense acomp worked/VBD Possible fine/JJ Opinion VBN Verb, past participle acomp proved/VBN Possible Verb, 3 rd person VBZ reliable/JJ Opinion singuar present acomp works/VBZ Possible JJ Adjective well/JJ Opinion
Selected Typed Dependencies xcomp – Open Clausal Complement Definition: An open clausal complement (xcomp)of a VP or an ADJP is a clausal complement without its own subject, whose reference is determined by an external subject. Example: Dependenc Component Indication Tag POS y Relation Example JJ Adjective xcomp easy/JJ Possible Opinion use/VB VB Verb, base form xcomp rendering/VBG Possible Opinion impossible/JJ VBG Verb, gerund or present participle xcomp found/VBD Possible Opinion VBD Verb, past tense difficult/JJ xcomp makes/VBZ Possible Opinion Verb, 3 rd person singular present VBZ ideal/JJ Verb, Non-3 rd person singular xcomp find/VBP Possible Opinion VBP convenient/JJ present xcomp experienced/VBN Not Opinion VBN Verb, past participle similar/JJ
Selected Typed Dependencies advmod – Adverbial Modifier Definition: An adverbial modifier(advmod) of a word is a (nonclausal) RB or ADVP that serves to modify the meaning of the word. Example: Dependency Component Indication Relation Example advmod well/JJ Possible Opinion amazingly/RB Tag POS advmod easily/RB Possible Opinion very/RB JJ Adjective advmod loads/VBD Possible Opinion fast/RB RB Adverb advmod looks/VBZ Not Opinion VBD Verb, past tense especially/RB advmod fits/VBZ Possible Opinion Verb, 3 rd person singular present VBZ perfectly/RB VBN Verb, past participle advmod recognized/VBN Not Opinion straight/RB advmod satisfied/VBN Possible Opinion very/RB advmod priced/VBN Possible Opinion reasonably/RB
Selected Typed Dependencies amod – Adjectival Modifier Definition: An adjectival modifier of an NP is any adjectival phrase that serves to modify the meaning of the NP. Examples: It has a nice design. amod (design, nice) – Opinion (design) I bought this new camera last year. amod (camera, new) – Not Opinion It has a pink cover. amod (cover, pink) – Not Opinion It looks nice. acomp (looks, nice) – Opinion (design)
Opinion Detection Algorithm 1. for each sentence in review text 2. set Opinion_Flag=False 3. check acomp_presence 4. if present 5. if governor is any form of verb 6. if dependent is any form of adjective 7. set Opinion_Flag=True 10. else if check xcomp_presence 11. if present 12. if governor is any form of adjective 13. if dependent is any form of verb 14. set Opinion_Flag=True 15. else if governor is any form of verb 16. if dependent is any form of adjective 17. set Opinion_Flag=True 18. else if check advmod_presence 19. if present 20. if dependent is any form of adverb 21. if governor in any form of verb 22. set Opinion_Flag=True 23. else if governor is any form of adverb 24. set Opinion_Flag=True 25. else if governor is any form of adjective 26. set Opinion_Flag=True
Pre-processed Data • Manually annotated 50 reviews for training, 50 reviews for testing. • Product Domain: Electronics , Product Type: Hard disk . • Example: Product Features Opinion Sentence ‘It was incredibly easy to set up and use.’ Usability ‘I like its design and the fact that I only Design need one cable.’ ‘Works perfectly and is completely Performance reliable, no problem at all.’ ‘I found this product really useful for Portability transport as it is that small.’ ‘The speed and capacity of the Passport Speed drive are impressive.’ ‘A satisfying product.’ General
Product Feature Assignment Counting Frequent Words: Only words in the components of the typed dependency relations are counted. Function words are ignored. Lemmatization is used to ensure counting of only the base form of words. Word counts are done within product feature scopes.
Product Feature Assignment Let, Total number of review lines: N j Set of product features: p , p ,..., p P 1 2 Frequency of the word w at review line i associated with product feature, p j : w i , j Word frequency count, WC j for word w within p j product feature scope can be denoted by the following equation: N WC w j i , j i 1 For different values of j , word frequency of the same word w will be different because associated product feature p j will be different.
Product Feature Assignment Word synonyms are taken using Wordnet’s synsets. Factors: Words in synsets are not originally present in review line Context might be different Polysemous synonyms exists. Let, for word w, Number of synsets : k Number of synonyms in i th synset : n i The probability of each synonym to be the appropriate synonym of the original word, w is considered by the following probability function: 1 P ( w ) k n i i 1
Recommend
More recommend